perceptron_Online.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Notebook Instructions\n",
    "<i>You can run the notebook document sequentially (one cell at a time) by pressing <b> shift + enter</b>. While a cell is running, a [*] will display on the left. When it has been run, a number will display indicating the order in which it was run in the notebook [8].</i>\n",
    "\n",
    "<i>Enter edit mode by pressing <b>`Enter`</b> or using the mouse to click on a cell's editor area. Edit mode is indicated by a green cell border and a prompt showing in the editor area.</i>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Neural Network\n",
    "\n",
    "In this notebook, we will use sklearn's MLPClassifier to create a trading strategy using the neural network. We will\n",
    "follow the below steps.\n",
    "\n",
    "1. Import the data\n",
    "2. Define predictors and target variables\n",
    "3. Split the data into train and test dataset\n",
    "4. Scale the data\n",
    "5. Create a neural network using the train data\n",
    "6. Analyze the performance"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Import the data\n",
    "\n",
    "We import the data of Boeing Co (ticker: BA) from the csv file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Open</th>\n",
       "      <th>High</th>\n",
       "      <th>Low</th>\n",
       "      <th>Close</th>\n",
       "      <th>Volume</th>\n",
       "      <th>Dividend</th>\n",
       "      <th>Split</th>\n",
       "      <th>Adj_Open</th>\n",
       "      <th>Adj_High</th>\n",
       "      <th>Adj_Low</th>\n",
       "      <th>Adj_Close</th>\n",
       "      <th>Adj_Volume</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Date</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2018-04-11</th>\n",
       "      <td>331.60</td>\n",
       "      <td>337.57</td>\n",
       "      <td>326.31</td>\n",
       "      <td>327.36</td>\n",
       "      <td>4566253.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>329.960125</td>\n",
       "      <td>335.900601</td>\n",
       "      <td>324.696286</td>\n",
       "      <td>325.741093</td>\n",
       "      <td>4566253.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-04-12</th>\n",
       "      <td>331.98</td>\n",
       "      <td>338.35</td>\n",
       "      <td>330.46</td>\n",
       "      <td>337.48</td>\n",
       "      <td>4318083.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>330.338246</td>\n",
       "      <td>336.676744</td>\n",
       "      <td>328.825763</td>\n",
       "      <td>335.811046</td>\n",
       "      <td>4318083.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-04-13</th>\n",
       "      <td>338.99</td>\n",
       "      <td>339.67</td>\n",
       "      <td>326.60</td>\n",
       "      <td>329.28</td>\n",
       "      <td>4730972.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>337.313579</td>\n",
       "      <td>337.990216</td>\n",
       "      <td>324.984852</td>\n",
       "      <td>327.651598</td>\n",
       "      <td>4730972.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              Open    High     Low   Close     Volume  Dividend  Split  \\\n",
       "Date                                                                     \n",
       "2018-04-11  331.60  337.57  326.31  327.36  4566253.0       0.0    1.0   \n",
       "2018-04-12  331.98  338.35  330.46  337.48  4318083.0       0.0    1.0   \n",
       "2018-04-13  338.99  339.67  326.60  329.28  4730972.0       0.0    1.0   \n",
       "\n",
       "              Adj_Open    Adj_High     Adj_Low   Adj_Close  Adj_Volume  \n",
       "Date                                                                    \n",
       "2018-04-11  329.960125  335.900601  324.696286  325.741093   4566253.0  \n",
       "2018-04-12  330.338246  336.676744  328.825763  335.811046   4318083.0  \n",
       "2018-04-13  337.313579  337.990216  324.984852  327.651598   4730972.0  "
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "data = pd.read_csv('BA.csv',index_col=0)\n",
    "data.index = pd.to_datetime(data.index)\n",
    "data.tail(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = data.rename(columns={'Adj Close': 'Adj_Close'})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Define predictor variables and a target variable\n",
    "\n",
    "We define a list of predictors as listed below \n",
    "1. ret1 is the one-day returns, \n",
    "2. ret5 is the five-day returns, \n",
    "3. std5 is the five-day standard deviation, \n",
    "4. volume_by_adv20 is the volume divided by the 20-days average of daily volume and \n",
    "5. price differences (High - Low) and (Open - Close). <BR>\n",
    "\n",
    "Syntax:\n",
    "1. <font color='blue'>np.log(data[--]/data[--]/shift(1))</font> computes the log return from the immediately previous row by default.\n",
    "2. <font color='blue'>rolling_sum(dataframe.column_name, window=n)</font> computes the sum of previous n rows.\n",
    "3. <font color='blue'>rolling_std(dataframe.column_name, window=n)</font> computes the standard deviation of previous n rows."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Returns\n",
    "import numpy as np\n",
    "data['ret1'] = np.log(data['Adj_Close']/data['Adj_Close'].shift(1))\n",
    "#data['ret5'] = pd.rolling_sum(data.ret1, 5)\n",
    "data['ret5'] = data['ret1'].rolling(window=5).sum()\n",
    "\n",
    "# Standard Deviation\n",
    "#data['std5'] = pd.rolling_std(data.ret1, 5)\n",
    "data['std5'] = data['ret1'].rolling(window=5).std()\n",
    "\n",
    "# Volume by ADV20\n",
    "#data['volume_by_adv20'] = data.Adj_Volume/pd.rolling_mean(data.Adj_Volume,20)\n",
    "data['volume_by_adv20'] = data.Adj_Volume/data['Adj_Volume'].rolling(window=20).mean()\n",
    "\n",
    "# High - low\n",
    "data['H-L'] = data['Adj_High'] - data['Adj_Low']\n",
    "\n",
    "# Open - Close\n",
    "data['O-C'] = data['Adj_Close'] - data['Adj_Open']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We define the target variable which is the future one-day returns. <BR>\n",
    "\n",
    "Syntax:\n",
    "<font color='blue'>shift(periods=n)</font> shifts the values by n period(s). If the value of n is negative then, it shifts the values backward and vice-versa."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Future returns\n",
    "data['retFut1'] = data.ret1.shift(-1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We drop the NaN values and store the predictor variables in X and target variable in y. The target variable is \n",
    "- 1.0 if the one-day future returns are positive and \n",
    "- 0 if the returns are negative or zero."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define predictor variables (X) and a target variable (y)\n",
    "data = data.dropna()\n",
    "predictor_list = ['H-L','O-C','ret5', 'std5','volume_by_adv20']\n",
    "X = data[predictor_list]\n",
    "\n",
    "import numpy as np\n",
    "y = np.where(data.retFut1 > 0.0, 1.0, 0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Split the data into train and test dataset\n",
    "\n",
    "This step is required to verify if the neural network we will create is any good. We will split the dataset into two parts, first 80% of the dataset will be used to create the model and the remaining 20% will be used as a test dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Split the data into train and test dataset\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "train_length = int(len(data)*0.80)\n",
    "\n",
    "X_train = X[:train_length-1000]\n",
    "X_test =  X[train_length:]\n",
    "\n",
    "y_train = y[:train_length-1000]\n",
    "y_test = y[train_length:]    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train1 = X[train_length-1000:train_length] \n",
    "y_train1 = y[train_length-1000:train_length]\n",
    "\n",
    "X_test1 =  X_test[:150]\n",
    "X_test2 =  X_test[150:300]\n",
    "X_test_val =  X_test[300:]\n",
    "\n",
    "\n",
    "y_test1 = y_test[:150]\n",
    "y_test2 = y_test[150:300]\n",
    "y_test_val = y_test[300:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(214,)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test_val.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Open</th>\n",
       "      <th>High</th>\n",
       "      <th>Low</th>\n",
       "      <th>Close</th>\n",
       "      <th>Volume</th>\n",
       "      <th>Dividend</th>\n",
       "      <th>Split</th>\n",
       "      <th>Adj_Open</th>\n",
       "      <th>Adj_High</th>\n",
       "      <th>Adj_Low</th>\n",
       "      <th>Adj_Close</th>\n",
       "      <th>Adj_Volume</th>\n",
       "      <th>ret1</th>\n",
       "      <th>ret5</th>\n",
       "      <th>std5</th>\n",
       "      <th>volume_by_adv20</th>\n",
       "      <th>H-L</th>\n",
       "      <th>O-C</th>\n",
       "      <th>retFut1</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Date</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2008-01-30</th>\n",
       "      <td>80.00</td>\n",
       "      <td>84.87</td>\n",
       "      <td>79.69</td>\n",
       "      <td>82.8699</td>\n",
       "      <td>11331100.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>60.815679</td>\n",
       "      <td>64.517833</td>\n",
       "      <td>60.580018</td>\n",
       "      <td>62.997365</td>\n",
       "      <td>11331100.0</td>\n",
       "      <td>0.023317</td>\n",
       "      <td>0.079067</td>\n",
       "      <td>0.018629</td>\n",
       "      <td>1.289782</td>\n",
       "      <td>3.937815</td>\n",
       "      <td>2.181686</td>\n",
       "      <td>0.003735</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2008-01-31</th>\n",
       "      <td>81.76</td>\n",
       "      <td>83.84</td>\n",
       "      <td>80.80</td>\n",
       "      <td>83.1800</td>\n",
       "      <td>7887700.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>62.153624</td>\n",
       "      <td>63.734831</td>\n",
       "      <td>61.423836</td>\n",
       "      <td>63.233102</td>\n",
       "      <td>7887700.0</td>\n",
       "      <td>0.003735</td>\n",
       "      <td>0.069182</td>\n",
       "      <td>0.019427</td>\n",
       "      <td>0.879880</td>\n",
       "      <td>2.310996</td>\n",
       "      <td>1.079478</td>\n",
       "      <td>-0.005062</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2008-02-01</th>\n",
       "      <td>83.01</td>\n",
       "      <td>83.99</td>\n",
       "      <td>81.85</td>\n",
       "      <td>82.7600</td>\n",
       "      <td>6685900.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>63.103869</td>\n",
       "      <td>63.848861</td>\n",
       "      <td>62.222041</td>\n",
       "      <td>62.913820</td>\n",
       "      <td>6685900.0</td>\n",
       "      <td>-0.005062</td>\n",
       "      <td>0.071750</td>\n",
       "      <td>0.018740</td>\n",
       "      <td>0.732508</td>\n",
       "      <td>1.626819</td>\n",
       "      <td>-0.190049</td>\n",
       "      <td>0.001690</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2008-02-04</th>\n",
       "      <td>82.51</td>\n",
       "      <td>83.36</td>\n",
       "      <td>81.91</td>\n",
       "      <td>82.9000</td>\n",
       "      <td>3106300.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>62.723771</td>\n",
       "      <td>63.369937</td>\n",
       "      <td>62.267653</td>\n",
       "      <td>63.020247</td>\n",
       "      <td>3106300.0</td>\n",
       "      <td>0.001690</td>\n",
       "      <td>0.066068</td>\n",
       "      <td>0.019428</td>\n",
       "      <td>0.344644</td>\n",
       "      <td>1.102284</td>\n",
       "      <td>0.296476</td>\n",
       "      <td>-0.014703</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2008-02-05</th>\n",
       "      <td>81.61</td>\n",
       "      <td>82.78</td>\n",
       "      <td>81.35</td>\n",
       "      <td>81.6900</td>\n",
       "      <td>4691600.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>62.039594</td>\n",
       "      <td>62.929024</td>\n",
       "      <td>61.841943</td>\n",
       "      <td>62.100410</td>\n",
       "      <td>4691600.0</td>\n",
       "      <td>-0.014703</td>\n",
       "      <td>0.008976</td>\n",
       "      <td>0.014019</td>\n",
       "      <td>0.536210</td>\n",
       "      <td>1.087080</td>\n",
       "      <td>0.060816</td>\n",
       "      <td>-0.017037</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             Open   High    Low    Close      Volume  Dividend  Split  \\\n",
       "Date                                                                    \n",
       "2008-01-30  80.00  84.87  79.69  82.8699  11331100.0       0.0    1.0   \n",
       "2008-01-31  81.76  83.84  80.80  83.1800   7887700.0       0.0    1.0   \n",
       "2008-02-01  83.01  83.99  81.85  82.7600   6685900.0       0.0    1.0   \n",
       "2008-02-04  82.51  83.36  81.91  82.9000   3106300.0       0.0    1.0   \n",
       "2008-02-05  81.61  82.78  81.35  81.6900   4691600.0       0.0    1.0   \n",
       "\n",
       "             Adj_Open   Adj_High    Adj_Low  Adj_Close  Adj_Volume      ret1  \\\n",
       "Date                                                                           \n",
       "2008-01-30  60.815679  64.517833  60.580018  62.997365  11331100.0  0.023317   \n",
       "2008-01-31  62.153624  63.734831  61.423836  63.233102   7887700.0  0.003735   \n",
       "2008-02-01  63.103869  63.848861  62.222041  62.913820   6685900.0 -0.005062   \n",
       "2008-02-04  62.723771  63.369937  62.267653  63.020247   3106300.0  0.001690   \n",
       "2008-02-05  62.039594  62.929024  61.841943  62.100410   4691600.0 -0.014703   \n",
       "\n",
       "                ret5      std5  volume_by_adv20       H-L       O-C   retFut1  \n",
       "Date                                                                           \n",
       "2008-01-30  0.079067  0.018629         1.289782  3.937815  2.181686  0.003735  \n",
       "2008-01-31  0.069182  0.019427         0.879880  2.310996  1.079478 -0.005062  \n",
       "2008-02-01  0.071750  0.018740         0.732508  1.626819 -0.190049  0.001690  \n",
       "2008-02-04  0.066068  0.019428         0.344644  1.102284  0.296476 -0.014703  \n",
       "2008-02-05  0.008976  0.014019         0.536210  1.087080  0.060816 -0.017037  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2569, 5)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data[predictor_list].shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Often, there is a regime change in time series data, so that a model trained on data randomly selected through the dataset will outperform a model trained on data selected from an earlier part of the data, but the latter will be more realistic. Also, if we are forced to select from an earlier part of the data, there is no way to ensure that the statistics of the train and test set are the same (i.e. equal % of classes in each), but that again is realistic. In trading applications, we normally don't have truly stationary statistics, and a ML strategy needs to be robust enough to withstand that."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Scale the data\n",
    "\n",
    "In this step, we standardize the features stored in X using sklearn's StandardScaler function. The machine learning model might behave in unexpected fashion if the individual predictors are not standardized.\n",
    "\n",
    "The StandardScaler function standardizes the predictors by removing the mean and scaling to unit variance."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# from sklearn.preprocessing import StandardScaler\n",
    "# scaler = StandardScaler()\n",
    "\n",
    "# # Create the scaler model using train data\n",
    "# scaler.fit(X_train)\n",
    "\n",
    "# # Transform the training and test data using the scaler model created above\n",
    "# X_train = scaler.transform(X_train)\n",
    "# X_test = scaler.transform(X_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create a neural network model\n",
    "\n",
    "The MLPClassifier function from the sklearn.neural_network package is used to create a neural network model. The MLPClassifier takes following as input parameter\n",
    "1. <b>activation:</b> The activation function such as logistic regression can be used to define the output from the neurons.\n",
    "2. <b>hidden_layer_sizes:</b> A tuple to specify the number of hidden layers and number of neurons in each of the hidden layers. For example, the tuple (2,5) indicates two hidden layers with 2 neurons in the first layer and 5 neurons in the second layer.\n",
    "3. <b>random_state:</b> It is used to initialize the weights and bias.\n",
    "4. <b>solver:</b> A function such as stochastic gradient descent is used to optimize the weights during backpropagation. 'sgd' is specified to indicate 'stochastic gradient descent'. For the full list of functions type help(MLPClassifier)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import the MLPClassifier\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "# Uncomment below line to see details of MLPClassifier\n",
    "# help(MLPClassifier)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# seed is initialized to 42\n",
    "seed = 42\n",
    "# Create the MLPClassifier model\n",
    "# mlp = MLPClassifier(activation='logistic', hidden_layer_sizes=(5), random_state=seed, solver='sgd')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Training the model\n",
    "\n",
    "The fit function is used to train the model using the train dataset which is stored in X_train and y_train. The parameters to the fit function are the predictor variables and target variable."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MLPClassifier(activation='logistic', alpha=0.0001, batch_size='auto',\n",
       "       beta_1=0.9, beta_2=0.999, early_stopping=False, epsilon=1e-08,\n",
       "       hidden_layer_sizes=5, learning_rate='constant',\n",
       "       learning_rate_init=0.001, max_iter=10000, momentum=0.9,\n",
       "       nesterovs_momentum=True, power_t=0.5, random_state=42, shuffle=True,\n",
       "       solver='sgd', tol=0.0001, validation_fraction=0.1, verbose=False,\n",
       "       warm_start=False)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Import the MLPClassifier\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "mlp = MLPClassifier(activation='logistic', hidden_layer_sizes=(5), random_state=42, solver='sgd',max_iter=10000)\n",
    "mlp.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Uncomment below line to see details of MLPClassifier\n",
    "# help(MLPClassifier)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Predict using the neural network\n",
    "\n",
    "We define a trading rule based on the predicted value from the neural network. If the predicted value is positive then we buy (+1) the stock and otherwise, we don't buy the stock. This signal is stored in predicted_signal. The strategy returns are generated by multiplying the future one-day returns by the predicted signal and stored in the strategy_returns_nn. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Open</th>\n",
       "      <th>High</th>\n",
       "      <th>Low</th>\n",
       "      <th>Close</th>\n",
       "      <th>Volume</th>\n",
       "      <th>Dividend</th>\n",
       "      <th>Split</th>\n",
       "      <th>Adj_Open</th>\n",
       "      <th>Adj_High</th>\n",
       "      <th>Adj_Low</th>\n",
       "      <th>...</th>\n",
       "      <th>Adj_Volume</th>\n",
       "      <th>ret1</th>\n",
       "      <th>ret5</th>\n",
       "      <th>std5</th>\n",
       "      <th>volume_by_adv20</th>\n",
       "      <th>H-L</th>\n",
       "      <th>O-C</th>\n",
       "      <th>retFut1</th>\n",
       "      <th>predicted_signal</th>\n",
       "      <th>strategy_returns_nn</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Date</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2018-04-10</th>\n",
       "      <td>331.35</td>\n",
       "      <td>336.69</td>\n",
       "      <td>329.55</td>\n",
       "      <td>334.83</td>\n",
       "      <td>5616231.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>329.711361</td>\n",
       "      <td>335.024953</td>\n",
       "      <td>327.920263</td>\n",
       "      <td>...</td>\n",
       "      <td>5616231.0</td>\n",
       "      <td>0.037582</td>\n",
       "      <td>0.012049</td>\n",
       "      <td>0.028752</td>\n",
       "      <td>0.873983</td>\n",
       "      <td>7.104690</td>\n",
       "      <td>3.462790</td>\n",
       "      <td>-0.022562</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-04-11</th>\n",
       "      <td>331.60</td>\n",
       "      <td>337.57</td>\n",
       "      <td>326.31</td>\n",
       "      <td>327.36</td>\n",
       "      <td>4566253.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>329.960125</td>\n",
       "      <td>335.900601</td>\n",
       "      <td>324.696286</td>\n",
       "      <td>...</td>\n",
       "      <td>4566253.0</td>\n",
       "      <td>-0.022562</td>\n",
       "      <td>-0.000244</td>\n",
       "      <td>0.030575</td>\n",
       "      <td>0.715461</td>\n",
       "      <td>11.204315</td>\n",
       "      <td>-4.219032</td>\n",
       "      <td>0.030446</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-04-12</th>\n",
       "      <td>331.98</td>\n",
       "      <td>338.35</td>\n",
       "      <td>330.46</td>\n",
       "      <td>337.48</td>\n",
       "      <td>4318083.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>330.338246</td>\n",
       "      <td>336.676744</td>\n",
       "      <td>328.825763</td>\n",
       "      <td>...</td>\n",
       "      <td>4318083.0</td>\n",
       "      <td>0.030446</td>\n",
       "      <td>0.003205</td>\n",
       "      <td>0.031367</td>\n",
       "      <td>0.726307</td>\n",
       "      <td>7.850981</td>\n",
       "      <td>5.472801</td>\n",
       "      <td>-0.024598</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3 rows × 21 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "              Open    High     Low   Close     Volume  Dividend  Split  \\\n",
       "Date                                                                     \n",
       "2018-04-10  331.35  336.69  329.55  334.83  5616231.0       0.0    1.0   \n",
       "2018-04-11  331.60  337.57  326.31  327.36  4566253.0       0.0    1.0   \n",
       "2018-04-12  331.98  338.35  330.46  337.48  4318083.0       0.0    1.0   \n",
       "\n",
       "              Adj_Open    Adj_High     Adj_Low         ...           \\\n",
       "Date                                                   ...            \n",
       "2018-04-10  329.711361  335.024953  327.920263         ...            \n",
       "2018-04-11  329.960125  335.900601  324.696286         ...            \n",
       "2018-04-12  330.338246  336.676744  328.825763         ...            \n",
       "\n",
       "            Adj_Volume      ret1      ret5      std5  volume_by_adv20  \\\n",
       "Date                                                                    \n",
       "2018-04-10   5616231.0  0.037582  0.012049  0.028752         0.873983   \n",
       "2018-04-11   4566253.0 -0.022562 -0.000244  0.030575         0.715461   \n",
       "2018-04-12   4318083.0  0.030446  0.003205  0.031367         0.726307   \n",
       "\n",
       "                  H-L       O-C   retFut1  predicted_signal  \\\n",
       "Date                                                          \n",
       "2018-04-10   7.104690  3.462790 -0.022562               0.0   \n",
       "2018-04-11  11.204315 -4.219032  0.030446               0.0   \n",
       "2018-04-12   7.850981  5.472801 -0.024598               0.0   \n",
       "\n",
       "            strategy_returns_nn  \n",
       "Date                             \n",
       "2018-04-10                 -0.0  \n",
       "2018-04-11                  0.0  \n",
       "2018-04-12                 -0.0  \n",
       "\n",
       "[3 rows x 21 columns]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['predicted_signal'] = mlp.predict(X)\n",
    "data['strategy_returns_nn'] = data.retFut1 * data.predicted_signal\n",
    "data.tail(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2228.0"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['predicted_signal'].sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Performance of the neural network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We calculate the Sharpe ratio for the strategy in train and test dataset. The risk-free rate is assumed to be 5% p.a."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# The below function returns the Sharpe ratio for the excess returns (Strategy returns - Risk free rate) passed to it.\n",
    "def annualized_sharpe_ratio(returns, N=252):\n",
    "    return np.sqrt(N) * returns.mean() / returns.std()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Sharpe ratio\n",
    "# Train dataset\n",
    "data['excess_daily_strategy_return'] = data.strategy_returns_nn - 0.05/252\n",
    "sharpe_train_nn = annualized_sharpe_ratio(data.excess_daily_strategy_return.iloc[:train_length-1000])\n",
    "# Test dataset\n",
    "sharpe_test_nn = annualized_sharpe_ratio(data.excess_daily_strategy_return.iloc[train_length+300:])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "def CAGR(returns):\n",
    "    cumulative_returns = returns.cumsum().iloc[-1]   \n",
    "    period_in_days = len(returns)\n",
    "    return 100*((cumulative_returns+1)**(252.0/period_in_days)-1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We compute the strategy CAGR in train and test dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# CAGR\n",
    "# Train dataset\n",
    "ret_train_nn = CAGR(data.strategy_returns_nn.iloc[:train_length-1000])\n",
    "# Test dataset\n",
    "ret_test_nn = CAGR(data.strategy_returns_nn.iloc[train_length+300:])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Plot cumulative returns in train dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import seaborn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the cumulative returns for neural network in train dataset\n",
    "data.strategy_returns_nn[:train_length-1000].cumsum().plot()\n",
    "\n",
    "# Show the legend\n",
    "plt.legend()\n",
    "\n",
    "# Show the graph\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Plot cumulative returns in test dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the cumulative returns for neural network in test dataset\n",
    "data.strategy_returns_nn[train_length+300:].cumsum().plot()\n",
    "\n",
    "# Show the legend\n",
    "plt.legend()\n",
    "\n",
    "# Show the graph\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Strategy Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: tabulate in c:\\users\\sundeep\\anaconda3\\lib\\site-packages (0.8.3)\n"
     ]
    }
   ],
   "source": [
    "!pip install tabulate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "|              |     Train |     Test |\n",
      "|--------------+-----------+----------|\n",
      "| CAGR         | -0.665247 | 28.7851  |\n",
      "| Sharpe Ratio | -0.163322 |  1.94475 |\n"
     ]
    }
   ],
   "source": [
    "from tabulate import tabulate\n",
    "\n",
    "print (tabulate(\n",
    "        [             \n",
    "            ['CAGR', ret_train_nn, ret_test_nn],\n",
    "            ['Sharpe Ratio', sharpe_train_nn, sharpe_test_nn],\n",
    "        ], \n",
    "        headers=['Train', 'Test'], \n",
    "    tablefmt='orgtbl'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Accuracy of the neural network classifier\n",
    "\n",
    "The classification_report function from sklearn.metrics package is used to analyze the performance of the neural network. The classification_report takes as input the actual output and the predicted output."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(214, 5)"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test_val.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "        0.0       0.43      0.77      0.55        88\n",
      "        1.0       0.64      0.28      0.39       126\n",
      "\n",
      "avg / total       0.55      0.48      0.45       214\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report\n",
    "print(classification_report(y_test_val, mlp.predict(X_test_val)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The output of classification report\n",
    "1. <B>Precision</B>: tp / (tp + fp) \n",
    "<BR>where tp is the number of true positives and fp the number of false positives. The precision is intuitively the ability of the classifier not to label as positive a sample that is negative.\n",
    "2. <B>Recall</B>: tp / (tp + fn) \n",
    "<BR> where tp is the number of true positives and fn the number of false negatives. The recall is intuitively the ability of the classifier to find all the positive samples.\n",
    "3. <B>F1-score</B>: It can be interpreted as a weighted harmonic mean of the precision and recall, The F1-score reaches its best value at 1 and worst score at 0. \n",
    "5. <B>Support</B>: It is the number of occurrences of each class in y_true."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Sundeep\\Anaconda3\\lib\\site-packages\\sklearn\\neural_network\\multilayer_perceptron.py:927: DeprecationWarning: The truth value of an empty array is ambiguous. Returning False, but in future this will result in an error. Use `array.size > 0` to check that an array is not empty.\n",
      "  if np.setdiff1d(classes, self.classes_, assume_unique=True):\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "MLPClassifier(activation='logistic', alpha=0.0001, batch_size='auto',\n",
       "       beta_1=0.9, beta_2=0.999, early_stopping=False, epsilon=1e-08,\n",
       "       hidden_layer_sizes=5, learning_rate='constant',\n",
       "       learning_rate_init=0.001, max_iter=10000, momentum=0.9,\n",
       "       nesterovs_momentum=True, power_t=0.5, random_state=42, shuffle=True,\n",
       "       solver='sgd', tol=0.0001, validation_fraction=0.1, verbose=False,\n",
       "       warm_start=False)"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mlp.partial_fit(X_train1, y_train1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Open</th>\n",
       "      <th>High</th>\n",
       "      <th>Low</th>\n",
       "      <th>Close</th>\n",
       "      <th>Volume</th>\n",
       "      <th>Dividend</th>\n",
       "      <th>Split</th>\n",
       "      <th>Adj_Open</th>\n",
       "      <th>Adj_High</th>\n",
       "      <th>Adj_Low</th>\n",
       "      <th>...</th>\n",
       "      <th>ret1</th>\n",
       "      <th>ret5</th>\n",
       "      <th>std5</th>\n",
       "      <th>volume_by_adv20</th>\n",
       "      <th>H-L</th>\n",
       "      <th>O-C</th>\n",
       "      <th>retFut1</th>\n",
       "      <th>predicted_signal</th>\n",
       "      <th>strategy_returns_nn</th>\n",
       "      <th>excess_daily_strategy_return</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Date</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2018-04-10</th>\n",
       "      <td>331.35</td>\n",
       "      <td>336.69</td>\n",
       "      <td>329.55</td>\n",
       "      <td>334.83</td>\n",
       "      <td>5616231.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>329.711361</td>\n",
       "      <td>335.024953</td>\n",
       "      <td>327.920263</td>\n",
       "      <td>...</td>\n",
       "      <td>0.037582</td>\n",
       "      <td>0.012049</td>\n",
       "      <td>0.028752</td>\n",
       "      <td>0.873983</td>\n",
       "      <td>7.104690</td>\n",
       "      <td>3.462790</td>\n",
       "      <td>-0.022562</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.0</td>\n",
       "      <td>-0.000198</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-04-11</th>\n",
       "      <td>331.60</td>\n",
       "      <td>337.57</td>\n",
       "      <td>326.31</td>\n",
       "      <td>327.36</td>\n",
       "      <td>4566253.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>329.960125</td>\n",
       "      <td>335.900601</td>\n",
       "      <td>324.696286</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.022562</td>\n",
       "      <td>-0.000244</td>\n",
       "      <td>0.030575</td>\n",
       "      <td>0.715461</td>\n",
       "      <td>11.204315</td>\n",
       "      <td>-4.219032</td>\n",
       "      <td>0.030446</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.000198</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-04-12</th>\n",
       "      <td>331.98</td>\n",
       "      <td>338.35</td>\n",
       "      <td>330.46</td>\n",
       "      <td>337.48</td>\n",
       "      <td>4318083.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>330.338246</td>\n",
       "      <td>336.676744</td>\n",
       "      <td>328.825763</td>\n",
       "      <td>...</td>\n",
       "      <td>0.030446</td>\n",
       "      <td>0.003205</td>\n",
       "      <td>0.031367</td>\n",
       "      <td>0.726307</td>\n",
       "      <td>7.850981</td>\n",
       "      <td>5.472801</td>\n",
       "      <td>-0.024598</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.0</td>\n",
       "      <td>-0.000198</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3 rows × 22 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "              Open    High     Low   Close     Volume  Dividend  Split  \\\n",
       "Date                                                                     \n",
       "2018-04-10  331.35  336.69  329.55  334.83  5616231.0       0.0    1.0   \n",
       "2018-04-11  331.60  337.57  326.31  327.36  4566253.0       0.0    1.0   \n",
       "2018-04-12  331.98  338.35  330.46  337.48  4318083.0       0.0    1.0   \n",
       "\n",
       "              Adj_Open    Adj_High     Adj_Low              ...               \\\n",
       "Date                                                        ...                \n",
       "2018-04-10  329.711361  335.024953  327.920263              ...                \n",
       "2018-04-11  329.960125  335.900601  324.696286              ...                \n",
       "2018-04-12  330.338246  336.676744  328.825763              ...                \n",
       "\n",
       "                ret1      ret5      std5  volume_by_adv20        H-L  \\\n",
       "Date                                                                   \n",
       "2018-04-10  0.037582  0.012049  0.028752         0.873983   7.104690   \n",
       "2018-04-11 -0.022562 -0.000244  0.030575         0.715461  11.204315   \n",
       "2018-04-12  0.030446  0.003205  0.031367         0.726307   7.850981   \n",
       "\n",
       "                 O-C   retFut1  predicted_signal  strategy_returns_nn  \\\n",
       "Date                                                                    \n",
       "2018-04-10  3.462790 -0.022562               0.0                 -0.0   \n",
       "2018-04-11 -4.219032  0.030446               0.0                  0.0   \n",
       "2018-04-12  5.472801 -0.024598               0.0                 -0.0   \n",
       "\n",
       "            excess_daily_strategy_return  \n",
       "Date                                      \n",
       "2018-04-10                     -0.000198  \n",
       "2018-04-11                     -0.000198  \n",
       "2018-04-12                     -0.000198  \n",
       "\n",
       "[3 rows x 22 columns]"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['predicted_signal'] = mlp.predict(X)\n",
    "data['strategy_returns_nn'] = data.retFut1 * data.predicted_signal\n",
    "data.tail(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2221.0"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['predicted_signal'].sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Performance of the neural network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We calculate the Sharpe ratio for the strategy in train and test dataset. The risk-free rate is assumed to be 5% p.a."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Sharpe ratio\n",
    "# Train dataset\n",
    "data['excess_daily_strategy_return'] = data.strategy_returns_nn - 0.05/252\n",
    "sharpe_train_nn = annualized_sharpe_ratio(data.excess_daily_strategy_return.iloc[train_length-1000:train_length])\n",
    "# Test dataset\n",
    "sharpe_test_nn = annualized_sharpe_ratio(data.excess_daily_strategy_return.iloc[train_length+300:])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Strategy CAGR\n",
    "CAGR represents the compounded annual returns of the strategy. The strategy CAGR is computed as follows:\n",
    "\n",
    "<i>(Cumulative returns + 1)^(252/number of days) - 1</i>\n",
    "\n",
    "where cumulative returns are the total strategy returns and number of days represent the total trading days."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We compute the strategy CAGR in train and test dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "# CAGR\n",
    "# Train dataset\n",
    "ret_train_nn = CAGR(data.strategy_returns_nn.iloc[train_length-1000:train_length])\n",
    "# Test dataset\n",
    "ret_test_nn = CAGR(data.strategy_returns_nn.iloc[train_length+300:])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Plot cumulative returns in train dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYEAAAEICAYAAAC55kg0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzt3Xd81PX9wPHX5y57QwYzEPaGAAFUQEAEFKzWgbjrrlq1VkvFat1bW0eLq1Wptmr9aetgKgqIiEDYewcIYSQhe9/d5/fH93uXu+zAJXdJ3s/Hgwd33/ve3Tvfu/u+v5+ttNYIIYRomyy+DkAIIYTvSBIQQog2TJKAEEK0YZIEhBCiDZMkIIQQbZgkASGEaMMkCQghRBsmSUAIIdowSQJCCNGGBfg6gPrExcXppKQkX4chhBAtyvr167O01vH17ef3SSApKYnU1FRfhyGEEC2KUupQQ/aT6iAhhGjDJAkIIUQb5tUkoJS6QCm1Wym1Tyk1p4bHuymllimlNiqltiilpnvz/YUQQjSO19oElFJWYC4wBUgH1imlvtJa73Db7RHgU631m0qpgcBCIKmx71VRUUF6ejqlpaVeiFy0NiEhIXTt2pXAwEBfhyKE3/Nmw/BoYJ/W+gCAUuoT4BLAPQloIMq8HQ1knM4bpaenExkZSVJSEkqpMwhZtDZaa7Kzs0lPT6dHjx6+DkcIv+fN6qAuwBG3++nmNnePA9cppdIxSgH3nM4blZaWEhsbKwlAVKOUIjY2VkqJQjSQN5NATWfkqsuWXQ3M01p3BaYDHyqlqsWglLpdKZWqlErNzMys+c0kAYhayHfDf5wsKGXPiQJfhyHq4M0kkA4kut3vSvXqnluATwG01quBECCu6gtprd/RWqdorVPi4+sd6yCE8BOlFXZeXrKbvOIK7A7N6Ge+Y+orP7ge11rz/a4TOByyrK2/8GYSWAf0UUr1UEoFAVcBX1XZ5zAwGUApNQAjCdR8qd8CvfrqqxQXFzf6efPmzSMj47SaR5pMWloaH330ka/DEC3Mf9Yd4W/L9jH62aUcOVX5W3Ce9P+38Sg3z0vlk3VHkPXN/YPXkoDW2gbcDSwBdmL0AtqulHpSKXWxudsDwG1Kqc3Ax8CNuhV9E+pKAna7vdbn+SoJ2Gy2Wh873SRQ198pWr8dGfkAlNkcPLtwp2t7zz8u5Mmvd/DN9hMA/PF/Wxn2xDfMW3XQJ3GKSl4dJ6C1Xqi17qu17qW1fsbc9qjW+ivz9g6t9Vit9TCtdbLW+htvvn9zKioqYsaMGQwbNozBgwfzxBNPkJGRwaRJk5g0aRIAERERPProo4wZM4bVq1fz5JNPMmrUKAYPHsztt9+O1prPPvuM1NRUrr32WpKTkykpKWH9+vVMmDCBkSNHMm3aNI4dOwbAunXrGDp0KGeffTazZ89m8ODBAIwfP55Nmza5Yhs7dixbtmypMe7HH3+c22+/nalTp3LDDTdgt9uZPXs2o0aNYujQobz99tsAzJkzh5UrV5KcnMwrr7zCvHnzuPvuu12vc9FFF7F8+fIa/86kpCQee+wxRowYwZAhQ9i1axcAK1asIDk5meTkZIYPH05BQc11xcuXL2fixIlcccUV9O/fn2uvvdZ11Vjbawvf+ml/FjPf+on/pFb2DflmxwkiQyo7IL636iDH8ysb7PNLbfxvk3+VgNsiv587qD5PfL3ddfXhLQM7R/HYLwbVuc/ixYvp3LkzCxYsACAvL4/333+fZcuWERdnNHMUFRUxePBgnnzySeN1Bw7k0UcfBeD6669n/vz5XHHFFfztb3/j5ZdfJiUlhYqKCu655x6+/PJL4uPj+c9//sPDDz/Me++9x0033cQ777zDOeecw5w5lWPxbr31VubNm8err77Knj17KCsrY+jQobXGvn79en788UdCQ0N55513iI6OZt26dZSVlTF27FimTp3K888/z8svv8z8+fMBo7RSm6p/J0BcXBwbNmzgjTfe4OWXX+Yf//gHL7/8MnPnzmXs2LEUFhYSEhJS62tu3LiR7du307lzZ8aOHcuqVasYN25cra8tfGv1/mzWpeVU235lSiI/7Mlk78lCALak53LLuB68+6NRAggLtJ7xe9/6z1SmDuzAlaMS699ZVCPTRpymIUOGsHTpUh588EFWrlxJdHR0tX2sViuXX3656/6yZcsYM2YMQ4YM4fvvv2f79u3VnrN79262bdvGlClTSE5O5umnnyY9PZ3c3FwKCgo455xzALjmmmtcz5k5cybz58+noqKC9957jxtvvLHO2C+++GJCQ0MB+Oabb/jggw9ITk5mzJgxZGdns3fv3kYdi6p/J8Bll10GwMiRI0lLSwOMEsr999/P66+/Tm5uLgEBtV+DjB49mq5du2KxWEhOTna9Rm2vLXyrtKKyGvCKkV2JiwgCoLjcRlBA5WnGoeG28T1d91cfyD6jRuKt6Xks3XmCP3xec8lX1K/FlwTqu2JvKn379mX9+vUsXLiQhx56iKlTp1bbJyQkBKvVuNIpLS3lrrvuIjU1lcTERB5//PEa+7JrrRk0aBCrV6/22J6TU/0qyyksLIwpU6bw5Zdf8umnn9Y762p4eLjH+/31r39l2rRpHvs4q3qcAgICcDgcrvvusbv/nU7BwcGAkSCcbQ9z5sxhxowZLFy4kLPOOoulS5fSv3//GmN0Pr/qa9T22sK3Sisqvxvr0k6RVVgOQHxEMO49doMCLHSM9iwBniouJy4imNNx6Rur3GKwExJopaC0giGPf8NrVyVzSXLVoUqiKikJnKaMjAzCwsK47rrr+P3vf8+GDRuIjIystZ7bedKMi4ujsLCQzz77zPWY+/P69etHZmamKwlUVFSwfft22rVrR2RkJD///DMAn3zyicfr33rrrdx7772MGjWK9u3bN/jvmDZtGm+++SYVFRUA7Nmzh6Kiomp/S1JSEps2bcLhcHDkyBHWrl3b4Pdw2r9/P0OGDOHBBx8kJSVF6vNbkdIKOzFhxjQd04d04n93ncOslETuO78vym0IUaCl+hiOQ9mN71EHxgWMza0U0f9Piym3OXhrxX4AHvrv1tN63bamxZcEfGXr1q3Mnj0bi8VCYGAgb775JqtXr+bCCy+kU6dOLFu2zGP/mJgYbrvtNoYMGUJSUhKjRo1yPXbjjTdyxx13EBoayurVq/nss8+49957ycvLw2azcd999zFo0CDeffddbrvtNsLDw5k4caJHFdTIkSOJioripptuatTfceutt5KWlsaIESPQWhMfH88XX3zB0KFDCQgIYNiwYdx4443cd9999OjRgyFDhjB48GBGjBjR6GP26quvsmzZMqxWKwMHDuTCCy9s9GsI/1Rqc9AuLIiVf5hEWFAAVotieLd2AB6Nww9M7QfA53eeTXZhOb/5aANzPt/C4vvOxWpR7DtZQI+4CKw1JIuqCsqMUuCFgzuyaNtxAPJKKpi7zEgCxeXe76lWWGYjIrh1nTaVv/fQTElJ0VWrN3bu3MmAAQN8FJHvFBYWEhERAcDzzz/PsWPHeO211wCjZDJx4kR27dqFxSIFvLb6HfGV2z5I5cipYhbfd261x7ak57Jo23EemNKXAKvnd/PP3+zmr9/vY/VD52Gza8a/uIwLB3fkjgm9GJYYU+d7HsouYsJLy3nh8iE8+Llx1f+HC/rx4uLdrn12PnkBoUFn3vgM8Mbyfby4eDcr/zCJxPZhXnnNpqSUWq+1TqlvPzlbtCALFiwgOTmZwYMHs3LlSh555BEAPvjgA8aMGcMzzzwjCUD4RFEdV8hDu8bw4AX9qyUAgOHdjBP9sbxS8kqMKslF245zydxV1fadu2wfSXMWkFVYBsCpIqPdISEyhNvGG5MFuicA43VLGvw3ZOSWUFjm2ca0/lAO87cY3Vj/+VMaAONfXNaqRjy3rnJNKzdr1ixmzZpVbfsNN9zADTfc4LHt/fffd5USnMaOHcvcuXObNMbG2Lp1K9dff73HtuDgYNasWeOjiMTpKii1ER/Z+MbdmDCjF1FeSQUbDtXe+QHgpSXGCf7D1Yf43ZS+rD14CoB24UFMHtCBv6+sPvDsvD+vIPWR8+tteNZaM/Gl5ZTbHWx/YhrhwQHsOVHA5W/+BEDPuAhiQoM4kW8koEvfWMVD0wdwVs/Yxv3BfkiSQCt10003Nbp9oLkNGTLEY5CbaLnySyvoGR9e/45VOEsP3+08wb9+PgxAz/hwDmQW4XBothzNo1d8OJEhgYQEWiitcPDad3u5eWwPnltkdCxoFxZIYVllG8KTlwyib4dIHv7fVvZnFvHRmsPcO7lPnXFsPJJLud3o4bTnRAHxkcEecx5l5JZQarMTHxlMZkEZm9Pz+PWH69n8WPVegS1Ni6078Pe2DOE78t1oXifzSzmWW0qHqNoH/9XGmQS2HjUGfE4b1IFZKcagry1H8/jl3FU8PX8nNrsDrSHAbDD+15rKNdQ7RocQFVK5gNC0QR05q2cs/71rLICr+qgu+04Uum7bHJrHv/Icw3OquJy8kgom909wbQvzUluDr7XIJBASEkJ2drb82EU1zkVl6hqNLLzrjeX7sWvNtWO6Nfq5zp5Dm4/kAvD8ZUNpZ1YRfb/rJABfbc6g98OLKLM5uHNiL8CYpgKgV3w4wQFWjx5ICWa1VHRoID3iwsk22w7q4myPAKOtwVnV9Pmd5xAeZOWjNYfJLa6ge2xlaedYXikZuZ5tDqlpp7C3sPaCFlkd1LVrV9LT06ltrQHRtjmXlxRNT2vNx2sPM2NIJ48TZENFhgTSMy6cA1lFAIQGWYkKNa7q31i2D4ASt9HIgzob3aJX7csGYMG94wHjhP+78/syfUhHj/UkwoOtlNTQVfSn/Vms2J3JWb1imdQvgfzSyiTw9IId5JfauOe83ozs3o57J/dxVT2d0yuWZy4dzKtL95JZUMaqfVnMNEsuP+7N4rp31/DIjAHc6jYq2t+1yCQQGBgoSwcK4QfyS2yU2Rz1duesy2/P78NvPzHahoIDLLQPN0oCtipX1H+5chjnD0jw2BZizj2klOK351ev9w+yWii3Oaptf/yr7ew5UciKPZlGEnArCaTnlGC1KH4zqTcA5/aNdyWBwV2iGZYYw5UpifT/02KPgW7bM/IA2HeykJakRVYHCSF8b/2hUzz8hdE/v3P06Ve/XTC4o+u2UoqR3dtxl1nt4+7S4V08upk+/cvB9b52oNXiavB1yiuuYI/ZBuA8+btXB2kNI7u3cyUY96om5yC2QKuFxHahHMwucj2241i+629oSVpkSUAI4XuXv1k5v9W5fU9/BcDgACvBARbKzCt2q0Xxhwv6c/O4HmTklnDx34wxA1VPrn0SIup97aAACwWlnn3/Nxw2uqIO6BRFurnwTX6pjfAgK0Vm1ZF7Q3Ok22133WPDOeSWBJztGqeK6m+I9ieSBIQQjfb3Hw64bl88rDPhZziVwsLfjqe4zLPuPi4imLiIYP5z+1keJ/I3rh3Bwawi17QUdQmyWqgwSwJ2h2bRtmP862ejZ9G43rH8fWU+SXOM6eD7JES4pry+fETlxHORwQGMTmrPhH6eia5HXDjrD+WgtWbjkVzSzKqh3OIKWhJJAkKIRskpKucZc9WwB6b0ddWdn4le8bVf1Y+pMiBr+pBODX7dQLNNYNHWY2TklfLU/B2ux64a3c1jgNkDU/txx7/WA3Ch23tYLIpP7zi72mt3jw2jsMzGpJeXuxIAGPMotSSSBIQQjbI2zeg+eceEXtxTzyAsXwsKMNoE7vz3Bo/tN5zdnZ5xlb2ZHrygP5MHJDCxXzwXurVR1OWCwR154usdHglgUr94juVVnyLen0kSEEI0irN65bIR/j9Xf6DVQmaBZx19UICFW8b18GhjSIoNI9BqYd5Noxv82p2iQ0mKDXMlgT4JEUSGBHIwq6ieZ/oX6R0kRDOau2wfv/+/zb4O44zY7EbXzYAGTPfsaz3iwqpNKb3kvnOrjWnoHBN6Wq8/ZWAH1+2xveMICbSQll3saiRuCSQJCNGMXlqym8/Wp7NiT8sd6OgsCQTWMCuov7n7vD5se2IaQ7pUrr0RElgZ94huMYzoFnPa4xwenjGQSWaD8fkDOlBirrB2w3uNX3TJV6Q6SIhm1CEqmBP5ZXyx8SgTzqBbpS85B3EFWP2/JADG/ESdokPYetQYzBUSUDnnz2d3nHPGr//iFcPYe7KAs3vGkltSztebM1rUVNOSBIRoJqv3Z7umIi4obVndCJ02HM7hjeXGdA4NWf3LX3RpV1nd4xwEBkbPnzMVHxnsmkb7oqGdWXvwFF9uyjjj120ukgSEaAZFZTau/vvPrvtVBzC1FLPeXk2F2SYQ2IIWMOrazlgJbGjXaK+tNFabqJBACkor0Fq3iNHDLedTFKIFc66BCxAVElBtBauWwr0qpaVUBwF0NKe57nga0103VnCABYemxcwmKklAiCa2JT2X2Z9tJjTQyoFnpzOpf4JfJIEym52H/7eVR77Y2uBp2bvFVq6t2xIahp2ck9LlFNc/rfSZCgwwjkvVOYv8lVQHiTZLa81N89ZxzehuTB3UsAFCjfHlpqPY7JoTBaVoDS9cMRSLRRERHEChH1QHfbvjBP9eY6zmdc3o7gzsHOXxeE3VGe5LSLaELqJOAzpFAnDNaax50FjO5Fhh0xDU+OeXVtgJsKga12RuCi0nlQvhZRl5pSzfncldVUaTekNucTm//WQTD/zfZnKKygkNtHLxsM4ARIQEkF1Uzn2fbHRNZuYLWW6DqPZlek5/XFphZ9qrP/DSkl0e293n92lJDcMxYUGkPT+DS4c3/ToTQWY12emUBA5lF9H/T4uZ8fqP3g6rVlISEG1OatopsgrLsTmMH2mfDpFefw/nqlgAh08Vu6ojwJg1E+CLTRlsOZrH9w9M9Pr7N0SO20Rnh6qMcl287Th7ThSy50Qh153VnU7RoZzIL2XjkRzG9Y7j/AEJLaLR0xeCzOqgMlv1xWzq41xcZ/eJAq/GVBdJAqLNufYfayizObjuLKNqoFv70xstWtWRU8XEhAVSYdfsd7uyXrL9hMdiKNOHdGR/ZiELthwjOrTmaYqbw8mCUmLDgwgKsPDnb/ewLSOPt69PATyTWGpaDrnFJzhZUIbdoXn20iEebQPCk7M6aNwLy1hw7zjXamgNkZlfWTqrsDuapd1FkoBoc5zz1i82e+ws2X6CgtKKWueNb6jxLy6jW/swjuQUU7WdtYfbZGX9O0Yx95oRBFs3scZcy9YXfj5wisT2Yfx+aj+ue3cNS7afcD12LK+EQKuiwq655+ONru0DOkVJAqiH+4l79/GCRiWBRduOuW4XltpoF34ajQqNJG0Cok05lle5MHhWYWVPEW/N53P4VGUC6BBV2Yha09w0SXHhZOSVeKxve7rWpZ3ip31ZDd4/p6icg1lFTOgbz7g+ca7tdodGa83BrCLO6RVX7XnjesdW2yY8uSeBsKDGXWe7Pzcjr6SOPb3Hq0lAKXWBUmq3UmqfUmpOLftcqZTaoZTarpT6yJvvL0R9tqQbUwd0MU/Kt47rwTm9Yjly6sx+cDX1Cf/zzGRmjjQaImvqSZOS1A6tYf0ho3HYZnfw5vL9HMwqYtorP7C2llLC/sxCjpwq9tg2863VXPOPNdz3yUbSGjCLpXPxlORuxpw5f7poIADZhWUcOVVCVmE5Uwd14O83pLie8+VvxvLA1H71vnZb1zmmcixCY9sFSirsRJoL9OxppnYBryUBpZQVmAtcCAwErlZKDayyTx/gIWCs1noQcJ+33l+Ihth2NA+rRTH/nnHcO7kPd5/Xm6iQQFcj8T0fbyRpzgLWHzJOwDa7g23mnDN1Kanw/LH/cXp/xvWJ4/6pfRnfJ85jkRKn5MQYrBbFhkM5PPblNh78fCsvLN7FpJeXs/tEAX9btq/ac9YfOsXkP6/ghvfWkldSwXMLd/KXb/e4Hv9iUwZPL9hZb7zOE0xfs1H87J6xWBS8teIAG48YSWl4YjvGm6WE4AALwxJjPKZcEDUb2jWGJfedCxi9rBqjuNzOgM5RhAZaeXvFAXKKmn5cgzfbBEYD+7TWBwCUUp8AlwA73Pa5DZirtc4B0FqfrPYqQpyGQ9lFlFY46Nex5p4+9368kU7RIew5UUCfhAjahQdx/5S+gDHy1Tkp2tebjTlfLn9zNWnPz+DpBTuZ91MaP8yeVGtd+Kmics5+7juPbTFhRl1up+hQPrxlTI3PCwsKwO7Q/PX76id7qOzb7u6xr7YDcDCriI/WHOZtt2UenZbuPFHvlAVPLzB+ls4F4gd2jmJEt3bsOJZHVKhxWujTIYJAq4W3rhvp2iYaJi7C+Pwf/Hwrl43o2uAG3pJyO51jQvjzlcO4698b+GbHcWaNatqxDd78ZLsAR9zupwNVv/19AZRSqwAr8LjWerEXYxBtkMOhmfDScgDW/HEyHWqYGuAr8+QeFxFUbVH0AIvCZteUV1kW0Ln2LMCJgtJak8CBzEJXY7NTSXnjuwdWZbd7VjGV2xxsO5rvul9XCeXT1CO1njwOZBZSak557J4oggMtlFY4yCupIDI4wHXiuqCBK22JSu5tAd/uOMHfVx7gg5tH19v5oKjcRlhQgKudpjnmmPJmm0BNlx1VK0oDgD7AROBq4B9KqWoTeSulbldKpSqlUjMzW+6866J55JVUNqzWtMj3J2sPu25nFZZzdpU1awOsFmx2Bxm5tbcLvFPDFXfV93x1VrKry6dzYNiZKCr3PAHklhhVA7eM6wFQbU2C9Y+cz0VDjWqnBz+vfSqI4/nG8od3TOjlsT3IXI83r6SCKB92XW0NQoOs/O58o6R590cb2Hg4l1X7sut8Tl5xBYeyi0mIDCbcTCL5LSwJpAOJbve7AlXnU00HvtRaV2itDwK7MZKCB631O1rrFK11Snx8y5xzXTSfbLd60yvfXl3t8Tn/3epx/7z+CR73A62KCofmsNnY+uQlg6q9xon82teNzTWT0PBuMWx+bCppz89ocNe+R2YMICzIyvcPTKj2WFGZZ2kiz0w2g7sY0zs45x+6MqUr824aRWxEsCsJAGzPMEoN/92Q7tEryvm6M6q0UwQFGEngZH4ZsRFN3zWxtXNOUeHsMxBXzzF96H9bAKPXmNWcXqQ5phz3ZhJYB/RRSvVQSgUBVwFfVdnnC2ASgFIqDqN6qPZLLCEa4JRbEnAvFQAs3Gr0ux7SJRqrRfHC5UOIjQj22MdqUdgdmnXmAuoXDOrIPef1dj0+uX9Ctdd1+mB1Gv/dkA5UtgM0xq3je7LjyQvoGR/Bot+O59lLh7geK6oyyZxzjvo+CZEkug1we/GKYUzsZyQ292qIi/76I5kFZdz/6WZ+4zY1RrFZwggL9mzkDbQai7LvOp7vajAWpy8uIojQQPdZV+s+3Z7ILyMhMpirRhnX0teO6caIbu2aNEbwYpuA1tqmlLobWIJR3/+e1nq7UupJIFVr/ZX52FSl1A7ADszWWtddRhKiFg6HZuW+LE4VldW6j3NeoOvO6sYVIxNrnO8mwGKhwu4gPaeEjlEhJESFcP+Uvq4G28T2Yfx8IBu7Q1d7/qNfGg21FoWra9/pGtApil7xEWw8nMPKvVnVqoO+2HSUyf0TGNwlmoX3jmfI499Ue43IEM8YjppVXM42AKgsQURUiTcowMLR3BLKbQ4GdPKcTE40nlKKxPah7DlhdMetb2rpnOJyRiW1dyWLh6YPaPIYwcsjhrXWC4GFVbY96nZbA/eb/4Q4bVprzn1pGek5JTWesLTWrrrvWSmJzByZWOsqUoFWo2G4zGYnwjyJKqWYYVatjOjejnk/pbHxcA4pSe1dr//5hqOu14gJC/LKKlVBARZemjmMm95f6zGYbf2hU6TnlLjmIIoMCWTF7InV2kCGdIlmzoX9eX6RMfGbczxBp+gQPlydxobDuZTbHIQGWj3mMwKjG6izcbymnkmi8RLbhbmSgKOe6bpziyuICWv+thjp9yVapEPZxaTnGFe5u4/nezy27Wgev/+/zew6bvSFv2xElzpP0AFWCzaHg3Kbg+CAyiL73GtGAMZSkIFWxXe7TrqSwLc7TniMMh7Xu/ro2jMRERLomkwM4Me9RoHZOdgNoHtsON2rDOANsFq4Y0IvVxJwtnN0iA7hnZUHXIPiZqUkVuu2GOR2f6CUBLzCfbqQutYddjg0ucXl1RJzc5BpI0SL5D7VQtXf1uVv/uRKAAC9EiLqfK0AizFHTmmFwzUDpLvIkEAS24dxKLvypPzDXs+eOe4Nst6QEBnMyfwyVw8f5xXiww2sIvjj9P4A7DxmJMggq8Wj2+p9U6r1xyDMrXrodNo3RHVzLuzvum2vpSRgd2ge/mIbDu2b4y5JQLQ4qWmneHP5fo9tnaJD2PToFP5xQ0q1Pvux9VxdOUfBFpbZPEoC7jpGhbBw63FXL6FD2Z7TNrhf8XlDz/hwSirs/LDXmA/IOSK5oYuixIQaf7Oz5JBXUkFWYTmT+yewYvZEOkVXn8todA+jlNPOB1USrVWA1cIHN48GwFHL8gK7jxfwsdmNeXi3aj3mm5wkAdHiXPHWateavc6+7oM6RxMTFsT5Azt4tBG8MmtYvfPeO5NAXkkFQQE1T4vgHP5/1TvGYvFp2UWc3TOWs3q2J9Cq6NLOO9NRO80cmUiXmFAe/GwLZTY7ucVGlVRYAxdJd3bxdFYH/W+j0X5x2YiudI+tOWE5+6Z3rCFBiNMXbvbCqq0k4FzysnN0CMMTmz8JSJuAaNFuOLs704d09Fj2sF+HCHYey2f2tH4NWknKeWLNLS6nV3zNVUfOn+/BrCLKbQ6O5pRwaXIXfjelLxV2XWM10pkICrDwy+GdmbtsP1vS83hrxX6CAiwNXsilpllLATpGB9e4HSp7r5xpLyfhyWJ+ZrW1CTiTwPs3jfbJQj3yaYsWq314EJ2iQ6qd8BLMaSOqTgNRG2df7pziiloXeXH//e4+XoBDGw2zSimCAprmh+vsIz7zLWMAXH1dDN31ToggOMDiUTUWEmips995d3NajFmjEmvdRzSeKwnUVhIwx7m0C/dNNZxUB4kWxXki7BgVwuLfjq/xysk5ZcOEfg0bbe4+M2ZCVM1XylF91RnXAAAgAElEQVRu/e+3Zxj17LVdbXtL1YT04S2jG/zcQKuFRy7ymMSXefVcaXaOCWXfMxdy+cimX4e3LXGOLaktiTt7cjnbcZqblAREi1JhLt79i2GdXFf8VQ3uEk3a8zMa/JpDu0Yzukd7ymyOalNKOD0wtR/rD+VQXG53TUPhPm98UxjY2bObZk2LvNQl2OzyeU6vWJ67bEitbQHu6hvVKhrPWRK4/cP1JCfGEBcRzFO/HMS+k4X06xhJkdlry9tVig0lSUC0KM4kkBDpvRNw55hQPv312XXuk5wYw2d3nMP011e6ttU0W6k3hQUF8NpVyfz2k02n9fxOZpKa1C+hQQlANA33UeblNgdLd55g6U5jKc93rh8J4JrW3Bck7YsWxWZOrxxgbf4GNPcr8+6xYc2ywMolyV0AXKOXG2Nc7zgW/XY8N41N8nJUojHcC1cPz/Ac5/Hx2sNYVOXMsL4gJQHRolSYna19VW3ROTqE3h0iXX2/m8Oepy+scc6j+iilZA4gP+DeDpPYznNNip/2ZxMbEUy4D3tkSRIQLYqzJBDohXl6TsdPD01u9vf0VV2x8A6rWxLoFhvGD7Mn8e81h3j7hwOU2Rz8YuiZrz1xJuTbJVqUyuog+eqKlqFqKa5bbJjH6nbeHmjYWFISEC2Kszoo0AdtAkKcDufkhe5rC5zdM5ZPf302FXYHI7s3/ZoBdZEkIFoUV3WQlAREC5EQGcy0QR08Rq9bLMo1V5OvSRIQLYqzi2iAj9oEhGisQKuFt69P8XUYtZLLKdFiVNgdPPG1sZKXlASE8A75JYkW4+cD2axLywEqB0IJIc6MVAcJv3TkVDHztxyjsKyC2dOMhTncF3vv31H6vwvhDZIEhN/Ze6KAKa/84Lr/wJR+WCyKglJjgfQ3rx3hq9CEaHWkOkj4nR3HPNcMPmgu65hvlgTG923Y7KBCiPpJEhB+xzm/utPkP68gs6CMd344AEB4A1fXEkLUT5KA8DsnC8qqbbv+3TVkm8nBF6svCdFaSRIQfmfTkVzAWPbRWf+/63iBL0MSotWSJCD8SpnNzvpDOQxLjOH7ByYSVmV2xYYutC6EaBhJAsKvHMwqoszm4OaxSXSMDqlW/5/6yPk+ikyI1kmSgPArpRXGtBCR5pq+PeIqV8S6YFBHwoKkV7MQ3iRJQPiVcpuRBIKsRgkgNiKYCWaXUJlXXwjvk1+V8CuuJOB2wh/eLQaonDxOCOE9kgSEXym324GqScCYb3239BASwuskCQi/4iwJuC8ak9zVKAlk1jB+QAhxZqSVTfiVMjMJBLuVBKLDApk9rR9n9Yz1VVhCtFpeLQkopS5QSu1WSu1TSs2pY78rlFJaKeW/Ky0In6gwVw5zNgw7/WZSb58vwydEa+S1JKCUsgJzgQuBgcDVSqmBNewXCdwLrPHWe4vWo6DUmCQuLFgGhQnRHLxZEhgN7NNaH9BalwOfAJfUsN9TwItAqRffW/ihrMIycovL69/RzfH8UoKsFtqHBTVRVEIId95MAl2AI273081tLkqp4UCi1nq+F99X+KmUp5cy/KlvG/WcE3mlJEQFY5E1hIVoFt5MAjX9arXrQaUswCvAA/W+kFK3K6VSlVKpmZmZXgxRNDet69/H3fH8UjpGydKRQjQXbyaBdCDR7X5XIMPtfiQwGFiulEoDzgK+qqlxWGv9jtY6RWudEh8vC4i0JSfyy+gYLUlAiObizSSwDuijlOqhlAoCrgK+cj6otc7TWsdprZO01knAz8DFWutUL8YgfGTX8XwWbzvuul9YZmv0a2itOZ4nJQEhmpPXkoDW2gbcDSwBdgKfaq23K6WeVEpd7K33Ef7pgldXcse/1rvuL9iSUeN+Gbkltb7GtqP5lFTY6dsx0uvxCSFq5tXBYlrrhcDCKtserWXfid58b+E7zm6dTntPFPDg51sB6NY+jP2Zhazck8noHrFMf30lf7poILeM61HtdRZsPUaARTF1YIdmiVsIISOGxRnafCSXNHMheACb3cG6tBwAesaFcyCriMl/XgHAtWO6AfDU/B1cPTrRY1roCruDt1bsByBGuocK0WwkCYgzcsncVR73ez+8yHU7ISqYA1mVCeLjtYddtwc+uoTfTu7D76b0BWDVvqwmjlQIURNJAqLJxEd6NvA6qnQXfe27vaQkteNYXinrzdLD078c3FzhCSGQJCDOgK5nEIBzdTCAsb1jWbUvG4Do0EDySox2hOvfXeva5+JhnbnurO5NEKkQojYylbQ4LXtOFNDjoYV17hMWWDn/j0VVjiWsbXGYsb1lllAhmpskAdFoaw+eYuorP7ju/+rs7vz16uEe+/xmUi9udusB5JwG+u5JvRnaNbrG1w2wyNdRiOYm1UGiUewOzZVvr/bY1qdDJHERwa77N56TxOxp/T32+fW5PZk5sisJUSHkFpeT/GT1OYVG92jfNEELIWoll16iUY7mVB/sFREcwKAuUa774W7TQD88fQAPXtCfAKuFBHMkcExYEL0TIgD4/M5zjOcEWUlsH9aUoQshaiAlAdFgaw+eqlYKcIoKCSQ+MpjMgjKP/v+3nduzxv37dYhk38lCkmLD+OyOs0mIlKkihPAFSQKiwQ65DQp77BfGekFPfL2DnvHhQOU0smFB9S8I88IVQ7l5XA9iI4KJdatKEkI0L0kCosHce4S2CwvikuTOTBnYga7tjGoc58k/PKj+r1VEcIAsFymEH5AkIBrslLlK2CuzhnFJcmeUUq4EAPD4xYP4aX825w1I8FWIQohGkiQgGuz5RbsAuHBwJ5SqvobQxH4JTOwnCUCIlkR6B4lGCw6Qr40QrYX8mkWDOBeMv3lsjxpLAUKIlkmSgGiQXccLAJjQT5b7FKI1kSQgGmS3mQT6y6pfQrQqkgREvUrK7Tz21XYAEiKlT78QrYkkAVGvX7utHSztAUK0LpIERJ1sdgcr92YCsPT+c30cjRDC2yQJiDplF5WjNTxz6WB6J0h7gBCtjSQBUatym4N0c9bQeJnfR4hWSUYMi1pd9NeV7DlRCEC8NAgL0SpJSUDUypkAQJKAEK2VJAHRIHFSHSREqyRJQNRIu88bDYQE1r9GgBCi5ZEkIGpUUGZz3ZZ5/4VovaRhWNQor7gCgBevGMrMkV19HI0QoqlISUDUqLTCDkBooFVGCQvRikkSEDUqrXAA0hYgRGsn1UHCQ3G5jZJyO2U2oyQQEijXCUK0ZvILFx6mv7aSkU8vdZUEggOkJCBEa+bVJKCUukAptVsptU8pNaeGx+9XSu1QSm1RSn2nlOruzfcXZy4tuxiobBOQkoAQrZvXfuFKKSswF7gQGAhcrZQaWGW3jUCK1noo8BnworfeX5w5h6NybMCtH6QCECuDxIRo1bx5mTca2Ke1PqC1Lgc+AS5x30FrvUxrXWze/RmQvod+5Fh+abVtnaNDfBCJEKK5eDMJdAGOuN1PN7fV5hZgkRffX5yhb7cfr7ZNuocK0bp5MwnUdLbQNWxDKXUdkAK8VMvjtyulUpVSqZmZmV4MUTidLCjlm+3HOVlQyuhnlrIlPZf3VqV57HP/lL6+CU4I0Wy8mQTSgUS3+12BjKo7KaXOBx4GLtZal9X0Qlrrd7TWKVrrlPj4eC+GKJxufG8dt3+4nqU7TnKyoIzXv9tLRm4Jv57Qk67tQgGY0FeOvRCtnTeTwDqgj1Kqh1IqCLgK+Mp9B6XUcOBtjARw0ovvLRrp8Cmjaca5dOTR3FJsDk3HqBA+vu0srh6dyMDOUb4MUQjRDLyWBLTWNuBuYAmwE/hUa71dKfWkUupic7eXgAjg/5RSm5RSX9XycqKJOav6F20z2gF2HssHICEyhMT2YTx32VACrdI9VIjWzqsjhrXWC4GFVbY96nb7fG++nzh9BaW2GrcnREmXUCHaErnUa4PsDs/2+rCgylHBie3CmjscIYQPSRJog9KyiwCYMaQTAB/eMsb1WAcpCQjRpsgEcm3Q9gyj/v+uSb2Ye+0IAObdNIrSCruMCxCijZEk0AbtyMgn0KrokxDp2jaxX4IPIxJC+IpUB7VBu47n0zshkqAA+fiFaOvkLNAG5RRXEBcR5OswhBB+QJJAG1RcZiMiWGoChRCSBNqkojIb4ZIEhBBIEmhztNbkl9oID5IVw4QQkgTanANZRRSW2ejbMbL+nYUQrZ4kgTam0JwuopMsFiOEQJJAm1Nc7lw7WKqDhBCSBNqckgqjJBAWJA3DQghJAm2OsyQQKiUBIQSSBNqcvJIKACJCpCQghJAk0OYcyCwiJNBCpyhpGBZCSBJoc/adLKRnXAQWi8wWKoSQJNDm7DtZSO+ECF+HIYTwE5IEWrg1B7Ips9kbtO89H2/kaG4JPePDmzgqIURLIUmgBftm+3FmvfMz//r5cLXHtNa8unQPu48XuLZ9vTkDkIFiQohKkgRasE/WHQHghUW7eHXpHtJzil2PFZTZeHXpXma9sxqA6a+tdD12xcjE5g1UCOG3JAm0YHtOGFf55XYHry7dy+0frHc9lldsdAXNNf/fccxYUvLWcT2wSqOwEMIkSaCFsTs0i7Yeo7TCTmZBmcdj7m0DzvEAVbeHyRTSQgg3ckZoIRwOzYq9mZTbHNz57w017tMpOtR1+8ipyqqh/2446rodHRrYdEEKIVocSQItxKtL9/D69/vq3Kfc5nDd/mFvluv2Q//d6rp9/gBZUF4IUanNVAfZ7A6m/GUFSXMW8J911XvT+LtF247Xu8/atFP0fXgRT3y9ncXbjnHxsM4ej7cPD6J7rHQPFUJUajNJ4Ie9mew9WQjAm8v3+ziaxjvpVv8/e1o/1+0ecZ4n9XK7g/dXpZFTXMH5Azvw55nDALh4WGdWPXhe8wQrhGgx2kx1UEl5ZVVJYVnDBlfVZdwL3zOsawxzrx1xxq9Vm5P5pTy7cCeXjejq0dB7Vs9YxveJY+XeLDrHhHAwq6jG5/eOj2Bg5yguH9m1yWIUQrRsbSYJ2BxGEpg2qANLtp/A7tCn1VXS2SsnPaeE9JwS5no7UDcP/N9mVu7N4otNGR7be8dH8O6vRlFud/Dol9tqfX778KAmjE4I0Rq0meqgCrsGwKKME/+8n9Ia/Fyb3eFqdL3zX+sZ/+Iy12M5ReW8vWI/NrujtqfXavX+bG79Z6rHczcezmH+lgyKy22sOXDKY/+xvWOJDQ8iKjSAoAALEcEBPDJjINeM6caOJ6fx0IX9PaaEiAmTnkBCiLq1nZKAeaK9+7zeLNp2nC83HeWWcT3qfd78LRnc/dFGIoID2PbENJbtzvR4/Mb317I5PY/N6bkczyvlnRtSiIsIblBMD3+xlQOZRWw6kktKUnsAfv3hek4WlPHExYMotzt4+/qRjO8Tx9qDp5jQNx6lPEsv7cODePbSIcZzJ/TiurO6M+ixJYzvEydLSAoh6tV2SgIOoyQQHxnMbeN7sPt4AVrrep9390cbASgsszFv1cFqj29OzwNg4dbjbDicy4erDzU4pghz4NaP+yq7c5ph8vFaowdTYrswwoICmNgvoVoCqEl4cAAHn5vOh7eMaXAcQoi2y6tJQCl1gVJqt1Jqn1JqTg2PByul/mM+vkYpleTN96+LsyQQaLEQFxFMmc1BUXnjGogf/3pHjdvdp2b+YlPlwCyHo/YkY7M72GVO7rZ6f7Zru7OZYp/Zk6lLu9Bqz61PQ5KFEEKAF5OAUsoKzAUuBAYCVyulBlbZ7RYgR2vdG3gFeMFb71+X43mlrrV1A6yKjuYsmmm19KoBY3qGgtKKWh93N7FvPFMGdiAqJIBD2cUUldk4cqqYnn9cyPwtGTU+52BWEeU2B3ERwWw8nOua66fMbHuwOTSx4UEywlcI0aS8WRIYDezTWh/QWpcDnwCXVNnnEuCf5u3PgMmqiS9bK+wOznruO15ashuAQKuFsb3jUAqW7TpZ6/N+8dcfOf8vK1z3n79sSK37JsWF8/cbUnhkhpHzcorL2WJWEy3aWvMgL+eEbndO7EW53cGag0ZpwH2en4uGdmrInyiEEKfNm0mgC3DE7X66ua3GfbTWNiAPiPViDNU8s2Cnx/1Aq1Ed1CM2nC1H86rtfyCzkH6PLGLHsXxO5JfRt0MEFwzqWOfC7KN7GI26wYHG4SyzOcg3SxFBATUf4l3HCwi0Kib2iwcgv9TGoewiSisqewqd2ze+EX+pEEI0njeTQE1X9FUrxRuyD0qp25VSqUqp1MzMzBqe0jCFZbZqXUGdYwPG9GzPqn1ZFJbZPB7/evMxV5UMQGmFg6AAC2FBRk+bQZ2jqr1PT3PUbpDVOJzHcktdE7hlFZZxPK+02nP2HC+gV3wEceFGT6L8kgouev1Hj32ki6cQoql5MwmkA+6rlXQFqlaIu/ZRSgUA0cCpKvugtX5Ha52itU6Jjz/9q+GiKid4dzNTEikutzPiqW85WVB5kg4P9uxWefhUMWsPniI0MMCMzajCARiV1A6AAPPk7ywJXPfuGt4wp6ZYuTeLs577rtr7ZxWVkxAVQkRIABYF2UVlFFSJNzJEkoAQoml5c5zAOqCPUqoHcBS4Crimyj5fAb8CVgNXAN/rhvTTPA07MvJ5+IvK2TNvPCeJm8Ymue4PT4wBjJk3Rz/zHRv+NIX24UHkl1ZPHMfzSwkKMEoQ3WPD+MO0fjwwpS82h6bErYdRcEDD+uWvOZDNqaIykmLDsFoUSXHh7D9ZvZE6Vkb8CiGamNdKAmYd/93AEmAn8KnWertS6kml1MXmbu8CsUqpfcD9QLVupN5SWGZj4+FcAF6ZNYzHLx7kMYNm1fboLenGvlmFZcSGB/HejSmux/569XCSE9vxp4sG8uIVQ1FKEWC1EBJopZ3bibq8yqjh687q5rrtzHXrD+Uw652fOXKqhHBznECn6BAWb69sQJ49rR9rH55MbAMHnQkhxOny6ohhrfVCYGGVbY+63S4FZnrzPWvj3ne/Q1T9C6sXmZPKZReWERcRzHn9O9AzPpwDmUUkJ8Zgtah6RxiXmqWCcb3jmD2tH8MSY+gcE8qLi3cz863VXDOmG+5DB5yliJgwzyv+sb3jSIiUxeCFEE2v1U4b4T55WkQDllT8NPUIM4Z2IquwnNgI47nf3HcuGbmlJLYPa9B7Th7QgdvP7cldE3u5Tuw5ReUApB7KIfVQjmvf0T3a88DUvgA8dGF/DmcXs/VoHi9eMZRks6pKCCGaWpuYNiK8AUlgxR5j6cb1h3Jcc/8EWC10i21YAgCjO+gfpw/wuLK/cEjNff0/uHk0XdsZr921XRgf3TaGh6cP4LLhVXvVCiFE02kTSaAhJQGAZxcaYwr6doioZ8+GG9GtHQefm84Mt4FfFw3tVG1yt8iQQG47t6erp5EQQjSHVn3GuXdyH6D+xdXn3zMOqJxe+pJk716NK6V4+YphPP6Lgex66gJev2q4V19fCCFOV6ttEwD43fl9uHtS71pH7cZHBpNZUEb/jpEe2yPrGB18ukKDrNw4tv6pq4UQojm16iSglHL176/Jv24Zw6HsIgKsFib2i2e5uVZAQ6uPhBCipWvTZ7t+HSPpZ5YCgs3SQniQVerlhRBthpztTEHmaF+ZqkEI0ZZIEjA5SwJRoW26cCSEaGMkCZicq3s1ZHSxEEK0FpIETOf0MpY1+Ns1I3wciRBCNB/VRJN4ek1KSopOTU1t8vex2R2UVNilTUAI0SoopdZrrVPq208qwE0BVguR0itICNHGyFlPCCHaMEkCQgjRhkkSEEKINkySgBBCtGGSBIQQog2TJCCEEG2Y348TUEplAofq2CUOyGqmcBrLX2OTuBpH4mocf40L/De2poiru9Y6vr6d/D4J1EcpldqQARG+4K+xSVyNI3E1jr/GBf4bmy/jkuogIYRowyQJCCFEG9YaksA7vg6gDv4am8TVOBJX4/hrXOC/sfksrhbfJiCEEOL0tYaSgBBCiNPUIpKAUqr21eJ9SOJqPH+NTeJqHH+NC/w3Nn+Nq0UkAdymvPazA9lSjp8Q3uavv0nw39+lXx4zfz1YACilLlBKLQFeVkpdCqD9oBFDKTVDKTUfeEopNdbX8TiZx+tLjLj8qi+0UuoXSqlPgDlKqe6+jsfJX4+Znx8vv/tNgt//Lv3ymIEfJgFlCFJKvQz8CZgL7AZmKqX6+DY6UEqNBB4D3gC2AL9SSt1oPtbsx9M8XiFKqXnAI8C7QARwi1IqrrnjqYlS6nyMz3IextXQPUqpGeZjcsyq8NPj5be/SfDb36VfHzMnv0sC2lAOLAYmaK2/An4CKoCDPg3OcD6wUmu9EPgSOI7xI43WWjuau5hnHq9SMxbn8fovRs8vfxkefz4wX2u9GHgbiARuVkqFa60dzR1MCzhm/ni8/Pk3Cf75u/T3Ywb4URJQSt2rlPq7UupWAK31Uq21TSk1HeMH2hd4Vik1y9y/WT5Ut7huMzctAy5SSrXTWpdgfKj5wB/MuJulmGfG9bxS6krzff+ntbab9z8H+imlnlJKjWuOeOqKDePLf45SKkRrfRIoBazATb6My1+OmZ8fL7/7TVaJzR9/l355zGrjF0nALLZdg/FDvF4p9UelVG/z4SzgAq312cByjCuipOb4UKvEdZ1S6mEgDfgG+FAptRLoCTwPtFdKhTdDTEop9TtgFpAKPKGUulEp1cHc5SRwHjAFOAbcqJSqdxKpJoztV8AuM5ZPlVLLgCiMq7Wo5iiq++sx89fjZcZ2I374m6whNr/4XdYQl18ds7r4y0Lzk4EXtNaLlVJZwEXAtcATWuu1bvvtADIBm4/iugS4QWt9j1KqG9BPa/2tUmoCEKK1LmrqgLTWWik1CXhEa71MKVUITAM08E+t9XLnvkqpLcA4oKSp46oltiJgKlAM3AoMBTprrRcopa4FejRH9Ya/HjN/PV4mf/1N1hSbz3+XtcTlT8esVj4tCbhd1WzEOGBorVOBn4FOqnoL/41AGJDto7hWAX2UUuO01oe11t+a+80A9jdlTFXiSgXGm3EtBvYAA5RSfas8ZRpGVUKTn9BqiW2RGVsK0EtrvVFrvcDcbySwxkdx+fyYtYDj5Ve/yXpi85ffpd8ds4Zo1iRQtf7L7apmFWBRSp1r3t+GURzubD7vBqXUNqAHcKdZ5+eruDKATubzzlVKrQD6AG95Mybz9a21xLUPiFRKDTHvrwCiMaoLgpRS15tXtN2BOVpru49jizL/oZSarpRaa8b2eRPEFe0en78cs0bG1ZzHq6P5v6VKXD79TZ5GbM35uxyklApx3venY3Y6mqt+cYxS6u/Ag+51rW4nkr3AdmCWUsqqtU4HOmIcLDC6fN2utf6V1vqEH8WVBtyltb7Um71KlFIpSqkPgUeVUr3ctjur79YCdmCKUipAa70D6AKMNHskHMH4kt1gNix6zRnENsp8fC9wh9b6cq11jpdisiilopTRR/x1AOdJ3O2zbPZjdoZxNdnxMt9/uFLqO+ApMy6HM2a3923236SXYkujaX6XQ5VSPwJPA7Fu231+zM5EkyYBpZRVKfUcxgx5q4ARwGPKbJBzu9oqAFYCQRgDKgKBdpgr7WitN2mtf/LDuA5rrbd7MS6LUupvGN0Cv8O4snlcKRWqlLJorW3m++4D1gG9gTnm08swV2DTWi/XWq/yVlxeii3NfHyv1nqDN2MzTxIFGJ9TF1XZ8yLA+Vn64pidYVxp5uNePV7K8ArwAUZ7yG1uj1ncrmqb9Tfp5di8+rt08wjwmZlcjppxWX15zLyhqUsCFowrrJla63nA74CzgFDnDkqpJ4CPgDzgUYyDttK8/8+2FJf5ZVoGTDbjegmj8dLudjX0lFLqXWA9xtXlaKXUeuAURu+IJuGF2JY0VWym/hiNba8B1yqlIp2JyVfH7AzjapLjpbXWGOMONmqtPzDj6OV+klVKPUXz/yb9NjbzAqgXUKi1ftXcNkUpFQMo8/7TzR2X12itvfoP42Ta17xtBWLM28Hm/18AKebtoRgHrpfb8y1AZFuMq8r284Fc4FvgZWAgcK4ZV2+3/SKcf0tbia3KZ+mcDj0QeB8YhHHCvQejHn2cxOX5OWK0N+zGOFmtwui//gFGibhvDXE1yXffn2OrIa5IjOqei8xzxRIzroeApOY8Zl7/W7140GKABRhFokeAiBr2iQQ2Y3R7q/qYpYm+ZC0lrnBzu/PkkQJMN28/BTwLdGvquPw5ttriMh87G3jNvH07xtX31+6ft8TlEde9wCaM5B0MvAA8B8T78jvm69jqieuPwAbgYvP+uRjjN85ujmPWVP+8WR0UjpEd7zFvj69hn9HAdq11hlIqQplzaCillG66/s8tJa5zoXJko9Y6VRtD4AEWYpx4T5lxWZowLn+Orca4TIcxetn8B2OU6AZgn9a6UOKqHpfW+nVgktb6B611GWZJGGOMgs++Y34QW12f5XyMq/725v1UjOkpSpshriZzRknA7PI0QSkVpY2GkneATzEOyhillLNrlLPnSDvgiFLqJozGsGTw/pDulh5XDUZgdIFzNiZ6/Yvmr7E1Iq52QDzGj3I4cAfGNBADJK6aP0ft2dtoJEY7mV98x5oztgbE1cV83y3AbOA3ypho8DpgCGZ//5aYAIDGLy+plFIY3Z4+AhwYgzHCgd9qszuWMgZHXAms01r/y+25H2KMoPsn8Ip5UL2itcWllIoCxmBUtRwHHtBa7/FWXP4cWyPjStVaf2hui3N7PAII0lqfkrhq/RyDMaqrXsboz+5P37Emje10P0tz+/0Y01L0AX6njW69LVajSgJmdyhnC/5RrfVk4C6MqgDXQsna6GqXBvRXRj/pCPOhBcCVWuubvHyibU1xRStj4rB8jN43T2utf9EEP06/jO004upnxhWutc5SRvdfi9a60Msn2tYSl/NzDDWrWsrxv+9Yk8Z2Bp9lpLn9Lxgn/2ktPQEADWsYxphj6FmMBpoJwC8w+vE6H1cY2XqC27YI4FWg7SkAAANiSURBVFWM6pUTQKeGvFdj/rXyuKo1Urfm2M4wrrUSl398jv4cm79+lr7+V29JQBmTMK3HqNvch9EbpAKYpJQaDa668yeBx92eOgMju24Chmitj9X3Xo3RBuLK8GZc/hybF+LaLHH5/nP059j89bP0Cw3InuOB693uvwHciTEJ0npzmwWjfu1TIMncdglwblNlL4mr9cQmcbWOuPw5Nn+Nyx/+NeTghWH01bWa968FnjNvbwLuMW+nAB83W+ASV6uJTeJqHXH5c2z+Gpc//Ku3OkhrXay1LtOV8+lMwRjwAsZKRwOUMUHWxxh9oJ0t701K4mo9sUlcrSMuf47NX+PyBw1eVEYZsx5qoAPwlbm5AGMU3WDgoDYnVdJmSm0OElfriU3iah1x+XNs/hqXLzWmi6gDYx6ULGComTX/BDi01j86D5wPSFytJzaJq3XE5c+x+WtcvtOYuiOMSZUcwI/ALb6uy5K4Wl9sElfriMufY/PXuHz1r1EjhpVSXYHrgb9oYzCHX5C4Gs9fY5O4Gsdf4wL/jc1f4/KVRk8bIYQQovXw6ULzQgghfEuSgBBCtGGSBIQQog2TJCCEEG2YJAEhhGjDJAkIUYVSyq6U2qSU2q6U2qyUul8pVedvRSmVpJS6prliFMJbJAkIUV2J1jpZaz0IY46Z6cBj9TwnCZAkIFocGScgRBVKqUKtdYTb/Z4Yi53EAd2BDzGWIgS4W2v9k1LqZ2AAcBBjmdLXgeeBiRizV87VWr/dbH+EEA0kSUCIKqomAXNbDtAfY7Ixh9a6VCnVB2Pa4RSl1ETg91rri8z9bwcStNZPK2O93FXATK31wWb9Y4SoR4NnERWijXNOKxwI/E0plQzYgb617D8VY4KyK8z70RgLk0sSEH5FkoAQ9TCrg+zASYy2gRPAMIw2tdLanoaxUMmSZglSiNMkDcNC1EEpFQ+8BfxNG3Wn0cAxrbUDYxIyq7lrARDp9tQlwJ1KqUDzdfoqpcIRws9ISUCI6kKVUpswqn5sGA3BfzEfewP4XCk1E1gGFJnbtwA2pdRmYB7wGkaPoQ3mClWZwC+b6w8QoqGkYVgIIdowqQ4SQog2TJKAEEK0YZIEhBCiDZMkIIQQbZgkASGEaMMkCQghRBsmSUAIIdowSQJCCNGG/T/xlG4PF4a7mAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the cumulative returns for neural network in train dataset\n",
    "data.strategy_returns_nn[train_length-1000:train_length].cumsum().plot()\n",
    "\n",
    "# Show the legend\n",
    "plt.legend()\n",
    "\n",
    "# Show the graph\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Plot cumulative returns in test dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the cumulative returns for neural network in test dataset\n",
    "data.strategy_returns_nn[train_length+300:].cumsum().plot()\n",
    "\n",
    "# Show the legend\n",
    "plt.legend()\n",
    "\n",
    "# Show the graph\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Strategy Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "|              |     Train |     Test |\n",
      "|--------------+-----------+----------|\n",
      "| CAGR         | 12.4311   | 30.1549  |\n",
      "| Sharpe Ratio |  0.498744 |  2.06808 |\n"
     ]
    }
   ],
   "source": [
    "from tabulate import tabulate\n",
    "\n",
    "print (tabulate(\n",
    "        [             \n",
    "            ['CAGR', ret_train_nn, ret_test_nn],\n",
    "            ['Sharpe Ratio', sharpe_train_nn, sharpe_test_nn],\n",
    "        ], \n",
    "        headers=['Train', 'Test'], \n",
    "    tablefmt='orgtbl'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Accuracy of the neural network classifier\n",
    "\n",
    "The classification_report function from sklearn.metrics package is used to analyze the performance of the neural network. The classification_report takes as input the actual output and the predicted output."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "        0.0       0.43      0.78      0.56        88\n",
      "        1.0       0.65      0.28      0.39       126\n",
      "\n",
      "avg / total       0.56      0.49      0.46       214\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report\n",
    "print(classification_report(y_test_val, mlp.predict(X_test_val)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Sundeep\\Anaconda3\\lib\\site-packages\\sklearn\\neural_network\\multilayer_perceptron.py:927: DeprecationWarning: The truth value of an empty array is ambiguous. Returning False, but in future this will result in an error. Use `array.size > 0` to check that an array is not empty.\n",
      "  if np.setdiff1d(classes, self.classes_, assume_unique=True):\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "MLPClassifier(activation='logistic', alpha=0.0001, batch_size='auto',\n",
       "       beta_1=0.9, beta_2=0.999, early_stopping=False, epsilon=1e-08,\n",
       "       hidden_layer_sizes=5, learning_rate='constant',\n",
       "       learning_rate_init=0.001, max_iter=10000, momentum=0.9,\n",
       "       nesterovs_momentum=True, power_t=0.5, random_state=42, shuffle=True,\n",
       "       solver='sgd', tol=0.0001, validation_fraction=0.1, verbose=False,\n",
       "       warm_start=False)"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mlp.partial_fit(X_test1, y_test1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Open</th>\n",
       "      <th>High</th>\n",
       "      <th>Low</th>\n",
       "      <th>Close</th>\n",
       "      <th>Volume</th>\n",
       "      <th>Dividend</th>\n",
       "      <th>Split</th>\n",
       "      <th>Adj_Open</th>\n",
       "      <th>Adj_High</th>\n",
       "      <th>Adj_Low</th>\n",
       "      <th>...</th>\n",
       "      <th>ret1</th>\n",
       "      <th>ret5</th>\n",
       "      <th>std5</th>\n",
       "      <th>volume_by_adv20</th>\n",
       "      <th>H-L</th>\n",
       "      <th>O-C</th>\n",
       "      <th>retFut1</th>\n",
       "      <th>predicted_signal</th>\n",
       "      <th>strategy_returns_nn</th>\n",
       "      <th>excess_daily_strategy_return</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Date</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2018-04-10</th>\n",
       "      <td>331.35</td>\n",
       "      <td>336.69</td>\n",
       "      <td>329.55</td>\n",
       "      <td>334.83</td>\n",
       "      <td>5616231.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>329.711361</td>\n",
       "      <td>335.024953</td>\n",
       "      <td>327.920263</td>\n",
       "      <td>...</td>\n",
       "      <td>0.037582</td>\n",
       "      <td>0.012049</td>\n",
       "      <td>0.028752</td>\n",
       "      <td>0.873983</td>\n",
       "      <td>7.104690</td>\n",
       "      <td>3.462790</td>\n",
       "      <td>-0.022562</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.0</td>\n",
       "      <td>-0.000198</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-04-11</th>\n",
       "      <td>331.60</td>\n",
       "      <td>337.57</td>\n",
       "      <td>326.31</td>\n",
       "      <td>327.36</td>\n",
       "      <td>4566253.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>329.960125</td>\n",
       "      <td>335.900601</td>\n",
       "      <td>324.696286</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.022562</td>\n",
       "      <td>-0.000244</td>\n",
       "      <td>0.030575</td>\n",
       "      <td>0.715461</td>\n",
       "      <td>11.204315</td>\n",
       "      <td>-4.219032</td>\n",
       "      <td>0.030446</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.000198</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-04-12</th>\n",
       "      <td>331.98</td>\n",
       "      <td>338.35</td>\n",
       "      <td>330.46</td>\n",
       "      <td>337.48</td>\n",
       "      <td>4318083.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>330.338246</td>\n",
       "      <td>336.676744</td>\n",
       "      <td>328.825763</td>\n",
       "      <td>...</td>\n",
       "      <td>0.030446</td>\n",
       "      <td>0.003205</td>\n",
       "      <td>0.031367</td>\n",
       "      <td>0.726307</td>\n",
       "      <td>7.850981</td>\n",
       "      <td>5.472801</td>\n",
       "      <td>-0.024598</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.0</td>\n",
       "      <td>-0.000198</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3 rows × 22 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "              Open    High     Low   Close     Volume  Dividend  Split  \\\n",
       "Date                                                                     \n",
       "2018-04-10  331.35  336.69  329.55  334.83  5616231.0       0.0    1.0   \n",
       "2018-04-11  331.60  337.57  326.31  327.36  4566253.0       0.0    1.0   \n",
       "2018-04-12  331.98  338.35  330.46  337.48  4318083.0       0.0    1.0   \n",
       "\n",
       "              Adj_Open    Adj_High     Adj_Low              ...               \\\n",
       "Date                                                        ...                \n",
       "2018-04-10  329.711361  335.024953  327.920263              ...                \n",
       "2018-04-11  329.960125  335.900601  324.696286              ...                \n",
       "2018-04-12  330.338246  336.676744  328.825763              ...                \n",
       "\n",
       "                ret1      ret5      std5  volume_by_adv20        H-L  \\\n",
       "Date                                                                   \n",
       "2018-04-10  0.037582  0.012049  0.028752         0.873983   7.104690   \n",
       "2018-04-11 -0.022562 -0.000244  0.030575         0.715461  11.204315   \n",
       "2018-04-12  0.030446  0.003205  0.031367         0.726307   7.850981   \n",
       "\n",
       "                 O-C   retFut1  predicted_signal  strategy_returns_nn  \\\n",
       "Date                                                                    \n",
       "2018-04-10  3.462790 -0.022562               0.0                 -0.0   \n",
       "2018-04-11 -4.219032  0.030446               0.0                  0.0   \n",
       "2018-04-12  5.472801 -0.024598               0.0                 -0.0   \n",
       "\n",
       "            excess_daily_strategy_return  \n",
       "Date                                      \n",
       "2018-04-10                     -0.000198  \n",
       "2018-04-11                     -0.000198  \n",
       "2018-04-12                     -0.000198  \n",
       "\n",
       "[3 rows x 22 columns]"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['predicted_signal'] = mlp.predict(X)\n",
    "data['strategy_returns_nn'] = data.retFut1 * data.predicted_signal\n",
    "data.tail(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2221.0"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['predicted_signal'].sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Performance of the neural network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We calculate the Sharpe ratio for the strategy in train and test dataset. The risk-free rate is assumed to be 5% p.a."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Sharpe ratio\n",
    "# Train dataset\n",
    "data['excess_daily_strategy_return'] = data.strategy_returns_nn - 0.05/252\n",
    "sharpe_train_nn = annualized_sharpe_ratio(data.excess_daily_strategy_return.iloc[:train_length+150])\n",
    "# Test dataset\n",
    "sharpe_test_nn = annualized_sharpe_ratio(data.excess_daily_strategy_return.iloc[train_length+300:])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Strategy CAGR\n",
    "CAGR represents the compounded annual returns of the strategy. The strategy CAGR is computed as follows:\n",
    "\n",
    "<i>(Cumulative returns + 1)^(252/number of days) - 1</i>\n",
    "\n",
    "where cumulative returns are the total strategy returns and number of days represent the total trading days."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We compute the strategy CAGR in train and test dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "# CAGR\n",
    "# Train dataset\n",
    "ret_train_nn = CAGR(data.strategy_returns_nn.iloc[:train_length+150])\n",
    "# Test dataset\n",
    "ret_test_nn = CAGR(data.strategy_returns_nn.iloc[train_length+300:])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Plot cumulative returns in train dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the cumulative returns for neural network in train dataset\n",
    "data.strategy_returns_nn[:train_length+150].cumsum().plot()\n",
    "\n",
    "# Show the legend\n",
    "plt.legend()\n",
    "\n",
    "# Show the graph\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Plot cumulative returns in test dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the cumulative returns for neural network in test dataset\n",
    "data.strategy_returns_nn[train_length+300:].cumsum().plot()\n",
    "\n",
    "# Show the legend\n",
    "plt.legend()\n",
    "\n",
    "# Show the graph\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Strategy Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "|              |     Train |     Test |\n",
      "|--------------+-----------+----------|\n",
      "| CAGR         | 6.04114   | 30.1549  |\n",
      "| Sharpe Ratio | 0.0962101 |  2.06808 |\n"
     ]
    }
   ],
   "source": [
    "from tabulate import tabulate\n",
    "\n",
    "print (tabulate(\n",
    "        [             \n",
    "            ['CAGR', ret_train_nn, ret_test_nn],\n",
    "            ['Sharpe Ratio', sharpe_train_nn, sharpe_test_nn],\n",
    "        ], \n",
    "        headers=['Train', 'Test'], \n",
    "    tablefmt='orgtbl'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Accuracy of the neural network classifier\n",
    "\n",
    "The classification_report function from sklearn.metrics package is used to analyze the performance of the neural network. The classification_report takes as input the actual output and the predicted output."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "        0.0       0.43      0.78      0.56        88\n",
      "        1.0       0.65      0.28      0.39       126\n",
      "\n",
      "avg / total       0.56      0.49      0.46       214\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report\n",
    "print(classification_report(y_test_val, mlp.predict(X_test_val)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Sundeep\\Anaconda3\\lib\\site-packages\\sklearn\\neural_network\\multilayer_perceptron.py:927: DeprecationWarning: The truth value of an empty array is ambiguous. Returning False, but in future this will result in an error. Use `array.size > 0` to check that an array is not empty.\n",
      "  if np.setdiff1d(classes, self.classes_, assume_unique=True):\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "MLPClassifier(activation='logistic', alpha=0.0001, batch_size='auto',\n",
       "       beta_1=0.9, beta_2=0.999, early_stopping=False, epsilon=1e-08,\n",
       "       hidden_layer_sizes=5, learning_rate='constant',\n",
       "       learning_rate_init=0.001, max_iter=10000, momentum=0.9,\n",
       "       nesterovs_momentum=True, power_t=0.5, random_state=42, shuffle=True,\n",
       "       solver='sgd', tol=0.0001, validation_fraction=0.1, verbose=False,\n",
       "       warm_start=False)"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mlp.partial_fit(X_test2, y_test2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Open</th>\n",
       "      <th>High</th>\n",
       "      <th>Low</th>\n",
       "      <th>Close</th>\n",
       "      <th>Volume</th>\n",
       "      <th>Dividend</th>\n",
       "      <th>Split</th>\n",
       "      <th>Adj_Open</th>\n",
       "      <th>Adj_High</th>\n",
       "      <th>Adj_Low</th>\n",
       "      <th>...</th>\n",
       "      <th>ret1</th>\n",
       "      <th>ret5</th>\n",
       "      <th>std5</th>\n",
       "      <th>volume_by_adv20</th>\n",
       "      <th>H-L</th>\n",
       "      <th>O-C</th>\n",
       "      <th>retFut1</th>\n",
       "      <th>predicted_signal</th>\n",
       "      <th>strategy_returns_nn</th>\n",
       "      <th>excess_daily_strategy_return</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Date</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2018-04-10</th>\n",
       "      <td>331.35</td>\n",
       "      <td>336.69</td>\n",
       "      <td>329.55</td>\n",
       "      <td>334.83</td>\n",
       "      <td>5616231.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>329.711361</td>\n",
       "      <td>335.024953</td>\n",
       "      <td>327.920263</td>\n",
       "      <td>...</td>\n",
       "      <td>0.037582</td>\n",
       "      <td>0.012049</td>\n",
       "      <td>0.028752</td>\n",
       "      <td>0.873983</td>\n",
       "      <td>7.104690</td>\n",
       "      <td>3.462790</td>\n",
       "      <td>-0.022562</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.0</td>\n",
       "      <td>-0.000198</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-04-11</th>\n",
       "      <td>331.60</td>\n",
       "      <td>337.57</td>\n",
       "      <td>326.31</td>\n",
       "      <td>327.36</td>\n",
       "      <td>4566253.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>329.960125</td>\n",
       "      <td>335.900601</td>\n",
       "      <td>324.696286</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.022562</td>\n",
       "      <td>-0.000244</td>\n",
       "      <td>0.030575</td>\n",
       "      <td>0.715461</td>\n",
       "      <td>11.204315</td>\n",
       "      <td>-4.219032</td>\n",
       "      <td>0.030446</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.000198</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-04-12</th>\n",
       "      <td>331.98</td>\n",
       "      <td>338.35</td>\n",
       "      <td>330.46</td>\n",
       "      <td>337.48</td>\n",
       "      <td>4318083.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>330.338246</td>\n",
       "      <td>336.676744</td>\n",
       "      <td>328.825763</td>\n",
       "      <td>...</td>\n",
       "      <td>0.030446</td>\n",
       "      <td>0.003205</td>\n",
       "      <td>0.031367</td>\n",
       "      <td>0.726307</td>\n",
       "      <td>7.850981</td>\n",
       "      <td>5.472801</td>\n",
       "      <td>-0.024598</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.0</td>\n",
       "      <td>-0.000198</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3 rows × 22 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "              Open    High     Low   Close     Volume  Dividend  Split  \\\n",
       "Date                                                                     \n",
       "2018-04-10  331.35  336.69  329.55  334.83  5616231.0       0.0    1.0   \n",
       "2018-04-11  331.60  337.57  326.31  327.36  4566253.0       0.0    1.0   \n",
       "2018-04-12  331.98  338.35  330.46  337.48  4318083.0       0.0    1.0   \n",
       "\n",
       "              Adj_Open    Adj_High     Adj_Low              ...               \\\n",
       "Date                                                        ...                \n",
       "2018-04-10  329.711361  335.024953  327.920263              ...                \n",
       "2018-04-11  329.960125  335.900601  324.696286              ...                \n",
       "2018-04-12  330.338246  336.676744  328.825763              ...                \n",
       "\n",
       "                ret1      ret5      std5  volume_by_adv20        H-L  \\\n",
       "Date                                                                   \n",
       "2018-04-10  0.037582  0.012049  0.028752         0.873983   7.104690   \n",
       "2018-04-11 -0.022562 -0.000244  0.030575         0.715461  11.204315   \n",
       "2018-04-12  0.030446  0.003205  0.031367         0.726307   7.850981   \n",
       "\n",
       "                 O-C   retFut1  predicted_signal  strategy_returns_nn  \\\n",
       "Date                                                                    \n",
       "2018-04-10  3.462790 -0.022562               0.0                 -0.0   \n",
       "2018-04-11 -4.219032  0.030446               0.0                  0.0   \n",
       "2018-04-12  5.472801 -0.024598               0.0                 -0.0   \n",
       "\n",
       "            excess_daily_strategy_return  \n",
       "Date                                      \n",
       "2018-04-10                     -0.000198  \n",
       "2018-04-11                     -0.000198  \n",
       "2018-04-12                     -0.000198  \n",
       "\n",
       "[3 rows x 22 columns]"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['predicted_signal'] = mlp.predict(X)\n",
    "data['strategy_returns_nn'] = data.retFut1 * data.predicted_signal\n",
    "data.tail(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2221.0"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['predicted_signal'].sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Performance of the neural network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We calculate the Sharpe ratio for the strategy in train and test dataset. The risk-free rate is assumed to be 5% p.a."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Sharpe ratio\n",
    "# Train dataset\n",
    "data['excess_daily_strategy_return'] = data.strategy_returns_nn - 0.05/252\n",
    "sharpe_train_nn = annualized_sharpe_ratio(data.excess_daily_strategy_return.iloc[:train_length+300])\n",
    "# Test dataset\n",
    "sharpe_test_nn = annualized_sharpe_ratio(data.excess_daily_strategy_return.iloc[train_length+300:])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Strategy CAGR\n",
    "CAGR represents the compounded annual returns of the strategy. The strategy CAGR is computed as follows:\n",
    "\n",
    "<i>(Cumulative returns + 1)^(252/number of days) - 1</i>\n",
    "\n",
    "where cumulative returns are the total strategy returns and number of days represent the total trading days."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We compute the strategy CAGR in train and test dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "# CAGR\n",
    "# Train dataset\n",
    "ret_train_nn = CAGR(data.strategy_returns_nn.iloc[:train_length+300])\n",
    "# Test dataset\n",
    "ret_test_nn = CAGR(data.strategy_returns_nn.iloc[train_length+300:])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Plot cumulative returns in train dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYYAAAEACAYAAAC3adEgAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzt3Xd4VFX6wPHvm95DSYBQg0hvAUJXAQUpFqxrF1B0V1dd110VG3aX3eWn6IqFdRG7srgqCorSFBGEUARCb0IAIXRCSD+/P+6dyUwyaWSSmSTv53nyZObce+ecmUzue0+554gxBqWUUsohwNcFUEop5V80MCillHKjgUEppZQbDQxKKaXcaGBQSinlRgODUkopNxoYlFJKudHAoJRSyo0GBqWUUm40MCillHIT5I0XEZHpwKXAIWNMFw/bBXgZGAVkAmONMavtbWOAx+1dnzPGvFNWfnFxcSYxMdEbRVdKqTpj1apVh40x8WXt55XAAMwAXgXeLWH7SKCt/dMXeB3oKyINgCeBZMAAq0RktjHmWGmZJSYmkpKS4qWiK6VU3SAiv5ZnP680JRljfgCOlrLLaOBdY1kO1BORBGA48J0x5qgdDL4DRnijTEoppc5OdfUxNAP2ujxPs9NKSldKKeUj1RUYxEOaKSW9+AuI3CkiKSKSkp6e7tXCKaWUKuStPoaypAEtXJ43B/bb6YOLpC/29ALGmGnANIDk5ORiwSM3N5e0tDSysrK8U2JVa4SFhdG8eXOCg4N9XRSlaoTqCgyzgXtE5GOszucTxpgDIjIPeEFE6tv7XQw8cjYZpKWlER0dTWJiItYgKKXAGMORI0dIS0ujdevWvi6OUjWCt4arfoR15R8nImlYI42CAYwxbwBzsYaqbscarjrO3nZURJ4FVtov9YwxprRO7BJlZWVpUFDFiAgNGzZEmx+Vv9t//AxNYsIICPD9OcwrgcEYc0MZ2w3wxxK2TQeme6McGhSUJ/q9UP7uRGYuAyYtBODd2/pwQbt4cvMLuPK1pWzYd5KnLuvE2IHVV+PVO5+r0JQpU8jMzKzwcTNmzGD//v1VUKKzt3v3bj788ENfF0OpWunEmVzn41unr+BIRjZz1x9gw76TADz15UbW7j3OoZPV04eqgaEKlRYY8vPzSzzOV4EhLy+vxG1nGxhKe59KKcuK3e4t6L2em8+fPl7rlnbF1KX0eWEBO9Mzqrw8Ghi85PTp01xyySV0796dLl268PTTT7N//36GDBnCkCFDAIiKimLixIn07duXZcuW8cwzz9C7d2+6dOnCnXfeiTGGWbNmkZKSwk033URSUhJnzpxh1apVDBo0iF69ejF8+HAOHDgAwMqVK+nWrRv9+/fnwQcfpEsXazaS888/n7VrC79UAwcOZN26dR7L/dRTT3HnnXdy8cUXc+utt5Kfn8+DDz5I79696datG2+++SYAEyZMYMmSJSQlJfHSSy8xY8YM7rnnHufrXHrppSxevNjj+0xMTOTJJ5+kZ8+edO3alc2bNwPw/fffk5SURFJSEj169ODUqVMey7h48WIGDx7MNddcQ4cOHbjpppuwWicp8bWVqknWpR0nJDCA+Q8MKrbt7sFt3J7Hhlf96LrqGpVUrZ7+MpWN+0969TU7NY3hycs6l7j9m2++oWnTpsyZMweAEydO8Pbbb7No0SLi4uIAK3h06dKFZ555xnrNTp2YOHEiALfccgtfffUV11xzDa+++iqTJ08mOTmZ3Nxc7r33Xr744gvi4+P55JNPeOyxx5g+fTrjxo1j2rRpDBgwgAkTJjjLMn78eGbMmMGUKVPYunUr2dnZdOvWrcSyr1q1ih9//JHw8HCmTZtGbGwsK1euJDs7m4EDB3LxxRczadIkJk+ezFdffQVYtZqSFH2fAHFxcaxevZrXXnuNyZMn89ZbbzF58mSmTp3KwIEDycjIICwsrMTXXLNmDampqTRt2pSBAweydOlSzjvvvBJfW6maIis3n3eX/Ur9iGDObRTF05d35rXF2zl4MpuxAxK5/bzWvLZ4B+0bRzP52u40jAqt8jJpjcFLunbtyvz583n44YdZsmQJsbGxxfYJDAzk6quvdj5ftGgRffv2pWvXrixcuJDU1NRix2zZsoUNGzYwbNgwkpKSeO6550hLS+P48eOcOnWKAQMGAHDjjTc6j7n22mv56quvyM3NZfr06YwdO7bUsl9++eWEh4cD8O233/Luu++SlJRE3759OXLkCNu2bavQZ1H0fQJcddVVAPTq1Yvdu3cDVk3mgQce4JVXXuH48eMEBZV8ndKnTx+aN29OQEAASUlJztco6bWVqinST2UDMLxzEwDGDEhk6cMXMumqrjx+SUcaRoWyduIwvrhnIF2bFz+vVIVaWWMo7cq+qrRr145Vq1Yxd+5cHnnkES6++OJi+4SFhREYGAhYw2vvvvtuUlJSaNGiBU899ZTHm/OMMXTu3Jlly5a5pR87VvI8gxEREQwbNowvvviCmTNnljnhYGRkpFt+//rXvxg+fLjbPo5mIoegoCAKCgqcz13L7vo+HUJDraucwMBAZ1/GhAkTuOSSS5g7dy79+vVj/vz5dOjQwWMZHccXfY2SXlupmuJ4ptXxfGGHRs60oMAAru/T0vm8XkRItZZJawxesn//fiIiIrj55pv561//yurVq4mOji6x3dxxIo2LiyMjI4NZs2Y5t7ke1759e9LT052BITc3l9TUVOrXr090dDTLly8H4OOPP3Z7/fHjx3PffffRu3dvGjRoUO73MXz4cF5//XVyc60v69atWzl9+nSx95KYmMjatWspKChg7969rFixotx5OOzYsYOuXbvy8MMPk5ycrP0Dqk7ac9QaoNIopuSm1OpWK2sMvrB+/XoefPBBAgICCA4O5vXXX2fZsmWMHDmShIQEFi1a5LZ/vXr1uOOOO+jatSuJiYn07t3buW3s2LH84Q9/IDw8nGXLljFr1izuu+8+Tpw4QV5eHvfffz+dO3fmP//5D3fccQeRkZEMHjzYrfmqV69exMTEMG7cuAq9j/Hjx7N792569uyJMYb4+Hg+//xzunXrRlBQEN27d2fs2LHcf//9tG7dmq5du9KlSxd69uxZ4c9sypQpLFq0iMDAQDp16sTIkSMr/BpK1XTzUn8jJCiAto2ifF0UJ3GM7qhJkpOTTdHmkU2bNtGxY0cflcg3MjIyiIqyvkyTJk3iwIEDvPzyy4BVgxk8eDCbN28mIEArhnXx+6H8nzGG1o/MJSE2jGWPXFTl+YnIKmNMcln76RmjBpszZw5JSUl06dKFJUuW8Pjj1kJ47777Ln379uX555/XoKCUnzpxJpfWj8wF4MAJ/5r8U5uSarDrrruO6667rlj6rbfeyq233uqW9vbbbztrEw4DBw5k6tSpVVrGili/fj233HKLW1poaCg///yzj0qkVNVxjEYCePn6JB+WpDgNDHXEuHHjKtzfUN26du3qdmOeUrVZVm7hrABDXEYk+QNtZ1BKKR/IyLaGVj8zujMxYf61VohXAoOIjBCRLSKyXUQmeNj+koistX+2ishxl235LttmV6YcNbEjXVU9/V4of/P0l6lcP80aat6zZf0y9q5+lW5KEpFAYCowDGtFtpUiMtsYs9GxjzHmzy773wv0cHmJM8aYSjewhYWFceTIERo2bKjTLCsnx0I9pU23oVR1STuWyZWv/eTWv9Am3n+GqTp4o4+hD7DdGLMTwF6lbTSwsYT9b8BayMermjdvTlpami7IoopxLO2plK9N/CLVLSjc3K8l4SGBpRzhG94IDM2AvS7P07CW7yxGRFoBrYGFLslhIpIC5AGTjDGfn00hgoODdelGpZTf+mn7YRZuPgTAyC5NeO2mnn7buuGNwODpnZXUqHs9MMsY4zpJf0tjzH4ROQdYKCLrjTE7imUicidwJ0DLli2LblZKKb+Was/4POmqrm7zIPkjb3Q+pwEtXJ43B0paZeZ64CPXBGPMfvv3TmAx7v0PrvtNM8YkG2OS4+PjK1tmpZSqVv9bs4+w4ACu692i7J19zBuBYSXQVkRai0gI1sm/2OgiEWkP1AeWuaTVF5FQ+3EcMJCS+yaUUqpGmr/xIJsOnCQrt8Bvm49cVbopyRiTJyL3APOAQGC6MSZVRJ4BUowxjiBxA/CxcR872BF4U0QKsILUJNfRTEopVdONfXsFi7dYg2IeG1Uz5uvyyp3Pxpi5wNwiaROLPH/Kw3E/AV29UQallPJHjqDQJCaMOy44x8elKR+dEkMp5XMrdx8lMyefNXuOMXZAIlm5BTSJrfn3nuTkFS5m9aehbX1YkorRwKCU8qnthzK49o3CFQqnzLeWkn3ysk6MG1hzh6BnZOdxxdSlzuedm8b4sDQVo3MlKaV8auiL33tMf/pLz92NxhhGv/ojq34teXlbf7Bu73G2H8oA4K1bk+nWvJ6PS1R+GhiUUj6TX1D6PFZncvKLpX28ci+/pJ3g6td/InHCHN5asrOqilcp2fmFzUgXtKtZQ+w1MCilfObkmVzn4xYNwott//Mnaxn+0g+cyMzFGMOhU1kcOH7GbZ/n5mwq8fWNMcxM2csLc0vep6q4BrXgQP8foupK+xiUUj5zwg4M/3dtd67u1ZxFWw7RrF4458RFcu5jX/NN6m8AXPTi9zx5WSfu/WiNx9c5cOIMCbFWYFmfdoIJ/1vHdb1bMPGLVOc+ASJMGNmhit9RoUw7MCx5aEiNuHfBldYYlFI+s8+++q8XYa1HMKR9I9o1jiYoMICo0MLr1sMZ2SUGBYDZawsnW/hu00FS9590CwoAb3xvzbTjOoldVdp1OIOgAKFhVEi15OdNGhiUUj7z1boDRIcG0e+chsW2uQaGkrx8fRIBAiezctl+6BSjXl7CrJS9Je7/2Zo0ej8/n0X2ZHZVZc66A0xdtIO8AkNESM1rmNHAoJTyic/X7OOjFXtIqBdGpIcg0L9N8WDhMPuegWx/fiSjk5oRGRrE6ex8Pluzj40HTrL/RJZzv3aNrbUOru1lTbs+6evNAGzYd8Kbb8VNRnYef/xwdZW9fnWoeaFMKVUr3P+Jtb7375I9Tyr31GWdiQ4L4qERHdiw7wTv/LSb3okN+CXtOF2axhIQYLXbhwQGMOOn3R5f45/XdGf8uyn8aWhb/rsqjYMnrWakoEDvXRM/8+VGpi/dRYsG4fRoUZ9LuiU4t3129wCv5VOdNDAopUpljOH9n/cwqksTGkaFeuU18+yhnOfERTL+fM/TRMRGBPPM6C4A9DunocfmJoAjp3OKpXVMiOFPF7Wle4t6rHxsaLHt69KOF0s7G1m5+UxfuguAvUfPsPfoGWb/YvV3DG4fTw8/XLazPLQpSSlVqt1HMnni8w1udydX1ml7xM4NXliX4E6X+Ydu6mu93qw/9GdElyZu+/3eZb+vN/xW6Xw//HkPHZ74psTtM8b1qXQevqKBQSlVqgdmWk0+Ow+f9tprjn17BQBhXljW8i8Xt3M+fvryzqx/6mKPfRaPjOrIV/eeR0IF52D69chpa+3wjGz2Hs1k1+HT7D2ayaOfrXfL19WyRy6s4LvwL9qUpFQdY4zhszX7GNGlSakjZvILDG8v3cWaPVazS3x0xZuR8gsMf/xgNc3rh/P4pZ04k5NPaFCA8zVdJ5k7W6FBhcElKDCA6FL6D7o0i+Wmvi2Z/O1WcvIKCAly3/fEmVzCggMIDQpkXdpxLn/VmuvomdGd3Ya/Pji8vfNxr1b1GTMgkb7nNGDElCUANI6u2RMAeqXGICIjRGSLiGwXkQketo8VkXQRWWv/jHfZNkZEttk/Y7xRHqWUZ1m5+dz9wWoemPkLLy/YVuq+/16y0+2u4vRT2Xy0Yk+F8ks7lsk3qb/x1o+7yMkr4KL/W8wd76Y4t7svz3L2nri0E6/d1LNc+zqC4dEifRP5BYbk577jlv9YtZnXFxeuMPz+8l/d9v3nvC0ALPjLID69y+pg7tAkhuBA4cIOjZwd4zVVpWsMIhIITAWGYS3zuVJEZntYcOcTY8w9RY5tADwJJGOtE73KPta/Z8dSqob68Oc9zvb11SVMQpdfYOjz/Hy3Tt0nL+vE019uZMr8rSX2C2Tm5AG41UJ2phc2P7V7/GsA53DSZvXCuaV/q0q8m0K3n1f+WVgD7ZP23R+s4n93D3Smn8rKJTffsGLXUcB95NLWgxkeX6tNfJTb883PjqRmhwSLN2oMfYDtxpidxpgc4GNgdDmPHQ58Z4w5ageD74ARXiiTUsqDzb+ddD7ekX6ap2ancux0DodOFY79P3kmt9hIn3EDW3NptwQOnszmyteWkjhhDou2HMIYwysLtvHBz7/SaeI8Ok2c5zyxQun9EtPH9nZrBqou4Xa/xob9J93SXd/zhn0n+PKXkpaut9x74bnF0gIDpMbXFsA7gaEZ4HqrYZqdVtTVIrJORGaJiGPgcnmPVUpVUE5eAX/8cDVz1h0A4C8zf2FmShoAcVGhHD2dw4yfdtPj2e/o8/wC5xDSN34obEK5Iqmps2PVMa+Ro3/g6dmpLNx8iBe/28pjn21wHvO7NwtHLz37ldVw8Ltk6wazsOAA4qNDeeHKrrRvEl0l77sso5OaAtbns3T7YWe6a9PSpf/60fn4ycs6OR/fNbiN83HHhJqzvkJFeSMweAqPRRsOvwQSjTHdgPnAOxU41tpR5E4RSRGRlPT09LMurFJ1xd5jmcxZd4A/friajOw8Pl1tBYWb+7Xk8u5Ni+3vWN9g+c7CK/5HR3VkzIBEoHBSOIdm9cPZVUKN4MXvtrLpQOEVecsGEQDEhgez8rGh3Ni38sNUz5ZrLeWmt352Pv58zb5i+17evSnjBrbm2dGdubRbgtvn1rVZbNUW1Ie8ERjSANdbF5sDbnUwY8wRY4xj5qp/A73Ke6zLa0wzxiQbY5Lj42vW3OZKVbes3Hy3Tt7XFm13Ph43sDXBQcWvyX47mcWMpbv4Za9VI9jxwigaxRSOrpkxrjf3DClsPgkPDmTBJs9zDr2yYBsjX7ZG6NzQp6Vz+GjDSO/cIFdZU65LKpZ2OMN9cr03bu7FKzf0AOCW/om8emNPOibEsOtvo9g96RJa2MGuNvLGcNWVQFsRaQ3sA64HbnTdQUQSjDEH7KeXA46hDvOAF0TEcXvgxcAjXiiTUnXaqwu3u3X8vmaPsNn0zAjCQwIJtTtWOyXE0KlpDLNWpbFy91HeX1446iiwSFt5dFgwfx3envsuakvv5+dTYGDX4dNc3bM5A9o0pE/rBqzZe5z7isyC2ikh2jks1NOaC77QrrF7M1ZefgHzUg/Ss2U9ZtzWh1W/HmNI+0Yej61pU2ifjUrXGIwxecA9WCf5TcBMY0yqiDwjIpfbu90nIqki8gtwHzDWPvYo8CxWcFkJPGOnKaXOUkZ2Hq+61BAcGkWHOjteg+3AEBESyORruwO4BYXuzUtuJgkJCqB3Yn1W7jrKbyezaFovjKt7NadFgwiPTVTXJrfgcIbVft+ivn9cZXdMcA8My3YeAeBYZi4xYcElBoW6wis3uBlj5gJzi6RNdHn8CCXUBIwx04Hp3iiHUgr2u6xwdu+F5/KvhVaQmPn7/s70YPsKPiy4+KigcxtFlTmdw419WzLfbkZybW4q6rkruhAWHEhMmHWqaeejDueiRIQrezRz9qvsPpIJwN+v7ubLYvkNvfNZqVrmgH2fwPW9W/CXi9vTuWksX67bT2JcpHOfOHsyvDbxVlqLBuHsPWoFlFdv7EH9yNIXl2nfpHBEzvDOjd22nRMf6WzGcsxddGPfVjSMCmVkkfmLfCl1/wn2HM0kccIcZ9q5jaJKOaLu0MCgVC1TYN9N/Lve1riOEV2aFJtQ7uqezUhqEUvLBlZg+Pb+QXScaE0I17IcnapNXeYbalRk+ofpY3oz6evNTLk+ydkeHxggjOqagD/xdNNagzICYl2hgUGpWiYv3woMwQEldyGKCOc2KmzWCQ8J5G9XdaVTQky5VhwTETY/O4K8guKjyxPjInnjll4ejvJvnm5Yq6s0MChVyzhuVAsKrNjomYpOge2pf6ImGdqxEfM3HWLxXwfz8cq9XO+FKcBrCw0MStUS2Xn55BcY/vSxNU12cAUDQ13z2k29yMzJo15ECBNGdvB1cfyKBgalaoEtv51i+JQf3NIc/QfKs5CgAEKCtE/BE12oR6laYMr8rW7PW8dFFltrQKny0m+OUrXAnqOZbs9LmsNIqfLQwKBULVA/wr1J5F/2HD9KnQ0NDErVAlm5+Qxo09D5fFinxqXsrVTpNDAoVQvk5FvrF4fbQ0hDSln3WKmy6KgkpWqBnLwCQgID+O6BC1j167FasYqY8h29rFCqBpm6aDtdnpxXLP1wRg4NIkNoXj+C0Um6CKKqHK8EBhEZISJbRGS7iEzwsP0BEdloL+25QERauWzLF5G19s9sb5RHqdrmSEY2G/ad4J/ztpCRnUeBy1QUufkFHDmdTeNSZjlVqiIq3ZQkIoHAVGAY1opsK0VktjFmo8tua4BkY0ymiNwF/AO4zt52xhhTfDklpZTT+f9Y5La05jVv/ESXZrE8fXlnNu4/iTHlm/xOqfLwRh9DH2C7MWYngIh8DIwGnIHBGLPIZf/lwM1eyFepOuGdn3YXW2959Z7jrN5znG9TD/LbSWua7T6tG/iieKoW8kZTUjNgr8vzNDutJLcDX7s8DxORFBFZLiJXeKE8StUaxhienJ1a4nZHUABoWs8/ls1UNZ83agyehj8Un4sXEJGbgWRgkEtyS2PMfhE5B1goIuuNMTs8HHsncCdAy5Y6C6KqG1o/MrfsnWxF12hW6mx5o8aQBrRwed4c2F90JxEZCjwGXG6MyXakG2P22793AosBj7dsGmOmGWOSjTHJ8fHxXii2Uv5t1a/Flz9v2SCCprFhPHFpJ5q51BCm3tizOoumajlv1BhWAm1FpDWwD7geuNF1BxHpAbwJjDDGHHJJrw9kGmOyRSQOGIjVMa1Unffp6n3F0lo1jOC92/sCMKprE2alpHFlz2Y0r68dz8p7Kh0YjDF5InIPMA8IBKYbY1JF5BkgxRgzG/gnEAX8117qb48x5nKgI/CmiBRg1V4mFRnNpFSd9eHPewD48eEhxEWF8sqCbVye1NS5PSE2nHsvauur4qlazCt3Phtj5gJzi6RNdHk8tITjfgK6eqMMStVWjtrAQyN0MRlVPfTOZ6X80I50a6H6do2jfFwSVRdpYFDKD02etwWAkV0SfFwSVRdpYFDKz/x2IouvN/xGg8gQ/jysna+Lo+ogDQxK+ZlXFm4D4OjpHB+XRNVVGhiU8jOOOY/OiYv0cUlUXaWBQSk/Y+x5Az64o69vC6LqLA0MSvmRt5bs5O/fbCZAoIlOo618RAODUn7kuTmbACgwYN8MqlS106U9lfIDi7YcYsGmg74uhlKABgalfO5kVi7j3l7pltaigU6hrXxHA4NSPnbIZU0FgIdHdODmfjq1vPId7WNQysdS9590e35Z9wSiw4J9VBqltMbg0fHMHOpFhPi6GKqO2HHImhdpzn3nsfdopk6hrXzOKzUGERkhIltEZLuITPCwPVREPrG3/ywiiS7bHrHTt4jIcG+UpzTzNx5kyOTFnCmyhq7D9kMZJD3zHZ+s3FPVRVEKsJbnbBITRuemsYzQuZGUH6h0YBCRQGAqMBLoBNwgIp2K7HY7cMwYcy7wEvB3+9hOWAv7dAZGAK/Zr1dl/jxzLbsOn6bjxG9YtuOI27Ytv51i6IvfA/DJyr2eDlfK606cySU2XJuOlP/wRo2hD7DdGLPTGJMDfAyMLrLPaOAd+/Es4CKxBmmPBj42xmQbY3YB2+3XqzLDOjZ2Pp6zvnAFUmMMw6f84Hx+LDO3KouhlNPJM3nEhGurrvIf3ggMzQDXy+s0O83jPsaYPOAE0LCcx3rNFVOX8r81hcsl9mpV3/m4aCBIalGvqoqhlJuTWVpjUP7FG4HB0+2Zppz7lOdY6wVE7hSRFBFJSU9Pr2ARLbn5BW7PM136GY5kZANw28DWAHy2Zp/ObqmqxcGTWTSMDPV1MZRy8kZgSANauDxvDuwvaR8RCQJigaPlPBYAY8w0Y0yyMSY5Pj7+rAr64PD2bs9nuvQjnDhj1RgGty987StfW3pW+SjlSW5+AV+vP4CxZ8k7k5PP8p1HOJyRoze0Kb/ijYbNlUBbEWkN7MPqTL6xyD6zgTHAMuAaYKExxojIbOBDEXkRaAq0BVZ4oUweNa9v/fM1jAzhyOkcfkk74dx2JteqPYSHBBIYIOQXGJrG6j+r8p7n52xixk+7AfhwfF9ufOtn57aGUVpjUP6j0jUGu8/gHmAesAmYaYxJFZFnRORye7f/AA1FZDvwADDBPjYVmAlsBL4B/miM8TyO1Aua1bPGh3drHltsm2P4anhwIJ/c2Q+ARjH6z+orOXkFZe9UQ+TmF/Dm9zucQQHgrR93ue1zQx+901n5D68MhTDGzAXmFkmb6PI4C7i2hGOfB573RjnKEh4SyJf3nEfLhhG8PH8b/00pbEo6YvcnhAUH0KVZLN2axzqbl1T1OZGZS/dnvgWs/p6JlxUd+Vz1nvxiAws2H2LBXwYRGlT50dP3fbSGrzf85pa2cPOhSr+uUlWlzk2J0bV5LLHhwYQGB5BtX5Wu+vUYj/xvPYDzRBAbHsy2gxk+K2ddtem3wukhpi/dxXqX5j5vyy8wbD14qlj6O8t+Je3YGZbvPOrxuM/X7GPS15vLlcdHK/YUCwpFPeWD4KdUaers4OnQoABy8gtInDCHprGFC6I4mo9EhP0nzlBQYAgI0Hnxq0tBgfugtMte/ZHdky7x2usbY3h32a88OTvVmTbv/gvYeOAEA9rE0dhlcZyPV+xhULvCwQhTF21n1+HTzFqVBkDK7qM8dklHerQsHPZ86GQWd32wmheu7Mq/Fm7jq3UHnNs6JsSw6YD7vEjefG9KeUudDQwhQYWVpf0nrNktY8ODnTWGpBb1+GFrOuc8Opf3b+/LeW3jfFLO6pSVm893Gw9yabcEny0Sk+lhqpLMnDwiQir/VU3df4JLXvmxWLrrjY0TRnYgNMiqTbrWJrYePMU/521xOy7l12Nc+dpPrHj0IqKShzR1AAAgAElEQVTDggkPCWTexoOs+vUYt81Yyb7jZwDru/b3q7tyefdmnDyTS0Z2HsNe+p5nRnep9HtSqirUuaYkB09tx65Xq5Ehhdu//MXjCNpa518Lt3HvR2tYvOXs7hPxhkx7dNj481o709bsOU5OnlW7e+P7HRV+TWMMg/65yGNQKGrS15udTYzHM3MxxnD0dA5XTC156HKfFxZw01vLrWPsvipHUAD4/O6BXNmjOYEBQv3IEFo0iGDzsyP5XXILj6+nlK/V2cAQE1b8CnTMgETn46DAwo9mR3rt7muYv/Egby3ZyamsPADGzVjpHGtf3e77aA0Adw1u40y76a2fWbHLau+fMn9rhV8zPSObX49kOp+fd24cjWNC2fnCKCJCSu5cPnI6h3mpv9Hz2e881mRcrd5znNz8AnYdPl1sW8eE6AqXWSlfqrOBoVPTGLfnA89tyAPD2jmfZ+UWngjS7buia6vx76bw3JxNhLo0r1X3Xd8nMnN5Ye4m5/OGUaHs+tso5/Ob/2ON+Y85i3UKDp8qfC8PDGvHe7f3YenDFxIQIIzqas1m+t7tfejuMg3KC1d2BeAP768u8XVv6dfK7fnri3e4TbnioGs3q5qmzvYxuHYyAnwwvp/b8yXbCptT6sqNbv9eUji2fs/RzGq76Sonr8A5RBXg94POATyfUEOCAkg7Vv41C05n53HbDGvZzPdu78P5ba3O5KBA67WfGd2ZCzs04vy28by9dDcAT1zaiet7t+DRz9YXe71nR3fm0m5NqR8ZwvyNB3lv+a/ObS9+V7w2M/e+88tVTqX8SZ2tMcRFhfLRHf2YfG13/jy0XbHtj4zs6HxcdI6l2mRd2nGP6VV1D8eprFxOnMklccIcEifMIf1UNgs2HXTbx/WkP6a/+1V52rEznPf3ReXOr/OT8/jNXjqzWb3iAT4iJMhZa2hij04LFIqNRLuqZzNWPHoRt/RPpH6ktYjTRR0bMWNcb1Y/Mcxt3yUPDWFI+3he/F33YjVTpWqCOltjAOjfpmGJ27q3qMfuSZcwZvoKjmfW3sn0dqZbbeLXJbfgE5cb/k5ne/8G9J93HuG6acvd0no/P7/Yfpd1K1ys5qERHSgwuF2Zn6246NJrQL+/4ByW7zjicbGcNvFRNCpSyxQRBrdvBMAvEy9m+JQfeO6KLrRoEMHb46p09nilqlSdrTGUV3BgALn5vumIrWr5BYb7P1kLwKOjOtKrVX1+f4HVjPPHD0tuWy9JXn4BK3YdZc2eY+xx6ex1KDoNhKs28ZF89+cL+PetyW7LqkaGBnFFj6YA9GhZ2Adw6FRWhcsXHVr6dVCrhpEs/OtgZ83h07v6O7ed2yiq1GNjI4JZ/uhFDO3UuNT9lKoJ6nSNoTyCA6XWNiUdOFE4pDImPIhP7xrAkYxs3vxhJ0CFb+4b+uL37HYJCEVv3jq3URTfbTxY9DAAWjSIoG3jaNo2Lj6Cp1erBvwy8WIWbjnImj1W09ej/9vAW2OSyyxTgMDN/Vrxl4vbV7gTuFerBs7Hul6Cqku0xlCG4MAA8goMh05mce9Ha8jIzvN1kbzG9b04TpqRLlfVJ7Mq1s+w20MtwdXriwvvQWgQGeK2zdMwT1exEcE0ii5syqkfUfKJOmX3UT5bk8ZDs36hwFj9SWd7Yn/8ko60bBChCzepOkUDQxmCAoWcvAIe/3wDX/6ynyVbfXfzl7c9+N91ANx3UVtnmuuQ1eMeljfNyStg5sq9bsN5AbYfKj7nUHZe4T5Fm35u7tuSnyZc6FxFrzx9GgPPjWPe/RfQKSHGOelhUXn5BVzzxjL+/MkvzEyxpq4IDjz7r/n488/hh4eGEBZcpUuRK+VXKhUYRKSBiHwnItvs3/U97JMkIstEJFVE1onIdS7bZojILhFZa/8kVaY8VSE0KJDsvAJS91tz3MTUoiaF9fusCepu7lc45bOIMH2s1URzzEOn++Rvt/DQp+vo8MQ3bukrdh0rtu8Jl8Cy/ZB1k2CLBuHcNrA1vx/Uhqb1wpl0lXW/QLSHGw49ad8kmvjoUA6XcG+JpxpdcKDeR6BURVS2xjABWGCMaQsssJ8XlQncaozpDIwApoiIa738QWNMkv2ztpLl8bqo0EBOZ+c5r37npf7GzJS97Dp82rmGQ02VbF+tuzbRAM7OX0eNYf7Gg0z7YQeZOXlMs/sfinIMOX368s7845puAM6pJaBwvYtXb+jJxMs6OZusHEM/r+9d/ukhGkaGcCTDc40hK7d4f5DrvFhKqbJVtvN5NDDYfvwOsBh42HUHY8xWl8f7ReQQEA94HkDvZ6JCgzmTm+9c4e3dZe7DJrc/P9Jt+oyaJDwk0G2kj0N9OzAcy8whKzef8e+mADCyyDDO/AJDoEvndGhQAGMGJDLbnlvqxJlctm0+yKB2jTh40rrCL9q3EBcVyoanh7vNTVWW0OAAcvMLyC8wvLxgG7f0a0W8PRTVtfnK+T61GUipCqnsGa2xMeYAgP27UWk7i0gfIARwnQntebuJ6SUR8bsl0+qV0skJnpsuaoqs3HzCPEwm6OjYPZaZy5DJi932d+V6El6w+ZCzhuDop/hszT5um5HCy/O3snTHYRpFh3q8ySwqNKhCI4aCAqzAsHznEV5ZsI2nvrSm0M4vMB5rNP3OKfl+FaVUcWUGBhGZLyIbPPyMrkhGIpIAvAeMM8Y46vuPAB2A3kADitQ2ihx/p4ikiEhKenr1dQD3ad2g1O2Oiedqmpkr97Jy9zEaRIUU2xYdFowInMjMIcPl/Tkmkmtsr1kx7m1rqomiV+mOjtr/2Pct/HdVGt9vSad1XKRX1rZw3Fvi2tQ1ed4W2jw6lw9+3lNs//gybmxTSrkrMzAYY4YaY7p4+PkCOGif8B0nfo/rFYpIDDAHeNwYs9zltQ8YSzbwNlDi7aLGmGnGmGRjTHJ8fHxJu3ldWZ3Nf/t6U6nb/UVWbj6ZOXnMTNnLos2HeOhTa0TSBR7WmQgMEKJDgziZlUf7JoX3FRy1O6M7JVjTPPy86yiHM7KdJ+hnR3cG3Ec2ARw4kUVGdh6tGpZvfqOyBAcJOfkFzg7o7LwCXl20vcT9i5ZHKVW6yvYxzAbGAJPs318U3UFEQoDPgHeNMf8tsi3BGHNArHaEK4ANlSyP1zUq42pz7vrSl230Fxe/9AN7jha/z6DoNA8OJ7Py3Bavh8Iagutss8nPFU5p4QiiJZ2IJ7jMP1UZIYEB5OUXlDoD7G0DW/PBz7+SnVegs5sqVUGVvZSaBAwTkW3AMPs5IpIsIm/Z+/wOuAAY62FY6gcish5YD8QBz1WyPF4XHBjA+qcupndi4Ujc2PBgrurZzIelqjhPQQFgSPtSu4U8enSU5xN8t+ZWR3ZJo7WKdjyfrey8AgoMziU2PZl4WScW/nWwzm6q1FmoVI3BGHMEuMhDegow3n78PvB+CcdfWJn8q0t0WDD//cMAEifMAWDipZ24uldz/rfamnt/3/EzHjtV/d2rN/Yo136dEmLY6LJW8ZGMHG4b2JrpS93nPmodFwlA52axzrRdfxvFu8t+9Tj66WylHbOCnOsqaa7mPzAIsGZTrYl/F6V8TRtfz4LjZqwb+1o3hh08WfEJ3aqbpykk+pdztE7RzttB7eNLvSEtNjyYKdcl8beruiIijBmQ6KxNeENpi8td1bNZmRPeKaVKp4HhLDjuabjUnsc/J8//J9lr4mGxoYiQkk/uE0Z2cD521AQcYsKC3Trl104cVmzCvCt6NOOGPi2pCqUFBr1nQanK08BwFhw3gDnuqK0Js6/meLjxq7TROn8Y1Ibx57UGICG2sIN62i29AOuOcAfXabKrQ0GRyLDysaHOx9FnsfSnUsqdTrtdAfMfGMRbS3ZyQTtruKxjcraaUGPwNFVEWfcUOFZxiwoL4r9/6E+zeuE0tdvsL+nWlG82/MZDIzqU9hJVIsclEPdoWY/46FB+mnAhU+Zv5cYqqqUoVZdojaECzm0UxaSruzmf15QaQ0GBIf1UNoPbxzunnihPp+wx+/6E1nGR9E5s4AwKYN2t/Pa4PnRMqP6lKx+/pJPzcYT9fprWC+cf13SnpZfulVCqLtMaQyU4agzZfl5jSM/IJie/gIs6Nubtsb2Zumg7wzs3KfO4Jy/rRIcm0fRt7V9TSrh2LpfWT6KUOjtaY6gERxu9Y0ppf7Ny91ESJ8zh/77dAkCzemGICPdc2NbjSmlFtWgQwV+Ht3ebKM9fPH9lF4AKTb6nlCofDQyV4OiU9b/TpuVPH60BcC5YU5uurrvbw18vLkfNRylVMRoYKsEx3fYrC7djShtD6SNZRZq4MnNq5oR/nnRpFsuGp4czqmtC2TsrpSpEA4OX7Egvfc3iqvTest3c/cGqYulFx/Tn5Plf8KqMqNDaUwNSyp9oYPAS14Xuq9sTX6R6nMzPdcqI2wa2ZmjHis+LpJSqezQwVNIt/VoB8Onqkid0q0qu6yp3nvgN7y3bTUGB4bYZ1kyoQzs2YuyARJ64tGONXWlOKVW9tC5eSVU5Yie/wJSZR/9JC5yPT+fk88QXqRzOyGHhZmtpjBv6tOSijo2rrIxKqdqnUpeQItJARL4TkW327/ol7JfvMuX2bJf01iLys338J/baDTVKon1D1TlF5hPyhsGTF3H9tGUet2Xl5nM4I9s5b5OrJdsKV7gb1K76FjVSStUOlW1bmAAsMMa0BRbYzz05Y4xJsn8ud0n/O/CSffwx4PZKlqfa3do/EYCerTzGxErZe/QMK3cf8zji6Q/vryL5ufkeJ5Rbvec4ADf0aaHNR0qpCqvsWWM08I79+B2sVdjKxV617UJg1tkc7y8CAoTEhhGcyc13Nv1426ns4sNMF28pe93rF67sWhXFUUrVcpUNDI2NMQfAWr8ZKGnYS5iIpIjIchFxnPwbAseNMY6zXhpQs5ZFs4kIc9YdoNtT8zhWynKTZ2vuugOlbv/vH/oXm/baUS6llKqoMgODiMwXkQ0efkZXIJ+Wxphk4EZgioi0wfMNwyVecovInXZwSUlPL/tquTo5Zlc9nZPPz7uOeuU1XZuPGsUUX3fatT862W7GurV/K2fau7f18Uo5lFJ1T5mjkowxQ0vaJiIHRSTBGHNARBKAQyW8xn77904RWQz0AD4F6olIkF1raA7sL6Uc04BpAMnJyX51p1aIy7oG2R7WPTgbrlNLR3qYyiIyJIhT2XmEBwc6awYJLovxNIyqcf34Sik/UdmmpNnAGPvxGOCLojuISH0RCbUfxwEDgY3GuiReBFxT2vE1QZDL5bu3ZlpNP5XtfLzJZb1lB8f00q6jksKCC/+cnoKJUkqVR2UDwyRgmIhsA4bZzxGRZBF5y96nI5AiIr9gBYJJxpiN9raHgQdEZDtWn8N/Klken3Ad+eOttRlmLN3tfPzUlxuLbU/dbwWLIe0Lh6O6ToERVcqazEopVZpKnT2MMUeAizykpwDj7cc/AR6HxxhjdgI1vjG8SUwom+z+YW+t5hZRzumkp97U0/m4fZPCqbQbVPNym0qp2kMHuXvBkA6Fg7G8VWMIt5uCBrWLd1uYxtXAcxu6TaXdxt4vLiq0zGU7lVKqJNre4AXBLk1J3qoxODqx60cEs/Nw4UJAy3Yc4YZ/LwegX5GV1WLCgnnj5l70bFXPK2VQStVNGhi8wLXzefK3W8nMyeehER0q9Zr5BYYAgfCQQLJzC4PN+8t/dT4O99DcNKKLLlyjlKocbUryguAi0068VokpuI+ezmHu+gPk5huCAgIIDQoky2XkUajLyKOwYF3WUinlfRoYvCAo0L09v1m9cLfn2w9l8Mj/1rmd4Ety9weruPuD1ew/fobAACE0KMBtCGzbRoUdzBoYlFJVQQODF/Sy7zy+98JzAUhq6d7G//v3UvhoxV62HcwodmxRB05kAdZ9DEEugcFxJ3SeS+d20RXalFLKG7SPwQsSYsOdcxV9m3qQ/Hz3G7NPZlnTQeXkl11jyLOPXbbzCCGBAYTaJ/+c/AJCgwLdbmgLD9G4rpTyPj2zeFlIUIDbdBZQOO/Rr0cyyzzedTlOKxhYfyJHc9KSbYed209ne2f6DaWUcqWBwctCggJYs+eYW1pMeDBgTbJXUc7AkFtAfoFh/b4TAFzTqzmD2+siPEop79PAUAWKjlKKi7RmRz2aUfaU3K43s3VvHktokNWUlJWbz9AXv3dum3xtd6LDgr1RXKWUcqOBwct6tarPyaxct7S8AqsZaO3eY4yY8gP3frTG46psAI1dptj+95hk541u5/9jEbsOnwbg0VGVu0dCKaVKo4HByyJCAsnKLWDj/pOMnrqU1XuOkWXfoLZoSzqbfzvFl7/sJ8PDqmwAacfOcGm3BHZPuoRG0WH8tONIsX3GDEisyreglKrjNDB4WVSoNdBr1CtL+GXvce77aA1ZHtZoyCyhv2H/8TM0q194H0TTIvdEAM7mJaWUqgqVCgwi0kBEvhORbfbv+h72GSIia11+shzLe4rIDBHZ5bItqTLl8QeRoe4jgNOOnWFn+uli+3mqMeTmF5Cbb4h2eY1B7dw7mP92la7jrJSqWpWtMUwAFhhj2gIL7OdujDGLjDFJxpgk4EIgE/jWZZcHHduNMWsrWR6fK+902Zkehpo67lFwvaP5gnbx3HF+a+fzEZ11LiSlVNWqbGAYDbxjP34HuKKM/a8BvjbGlD2gv4YqoU+5mB+2FV+3+sBx667nU1nutYmWDSKcj3UBHqVUVatsYGhsjDkAYP9uVMb+1wMfFUl7XkTWichLjiVAa7JjmWUPSQX457wtxdLeWbYbgA9X7HFLbxwT5nxcdCisUkp5W5lnGRGZLyIbPPyMrkhGIpKAtZLbPJfkR4AOQG+gAdZSnyUdf6eIpIhISnp68attf3Flj2Ye09vERwLw6V39Szy2fWNrgryJl3ZyS28QqauxKaWqT5ntEsaYoSVtE5GDIpJgjDlgn/gPlfJSvwM+M8Y4B/k7ahtAtoi8Dfy1lHJMA6YBJCcnl7PBpvrVK2FJzQ/v6Oe88g8LDmBUlwTnNmMML83fxisLtgHQqmGE27Et7KYk7V9QSlWHyjZYzwbGAJPs31+Usu8NWDUEJ5egIlj9ExsqWR6/9POjF7k1B8VHh1Jgd0Zk5ebzr4XbmLqocA2H/AL3uNc4JoyVjw2lodYclFLVoLIN1pOAYSKyDRhmP0dEkkXkLcdOIpIItAC+L3L8ByKyHlgPxAHPVbI8fimoyPrLIYEBZNijkt74fodbUABIalF8ac74aF3HWSlVPSpVYzDGHAEu8pCeAox3eb4bKNb4boy5sDL5+7u5953Pil1HaBjl3qfeq1V95qw7QH6B4eQZ9xFIQQGCVYFSSinf0CEuVSgxLoKxA1sXS+/cNJbTOfkcPZ2Dwb3ZSDualVK+poGhCpU0tDQ8pHDG1KL3PZT3BjmllKoqGhiqQMeEGKB434JDiB0wcvMLis2yqvMgKaV8TW+jrQIfju/L9vSMEvsKgp2BwVBkABL32OtGK6WUr2hgqAL1I0PoHdmgxO3BgVbAyM0v4JTL2g07XxilI4+UUj6nTUk+EGwv1zlrVZrbGs8aFJRS/kADgw84+hVm/LTb2acw9caeviySUko5aWDwgaOnC5uP1qUdp1vzWC7pllDKEUopVX00MPhA56Yxzscns/JYl3bCh6VRSil3Ghh8oGNCDE1jw5zLgCqllD/RwOAjocGBHpf3VEopX9PA4COOIatKKeVvNDD4iK7EppTyV3p28pEgl8DQOi7ShyVRSil3lQoMInKtiKSKSIGIJJey3wgR2SIi20Vkgkt6axH5WUS2icgnIlJnphYNsZuSerWqz7d/vsDHpVFKqUKVrTFsAK4CfihpBxEJBKYCI4FOwA0i4ljU+O/AS8aYtsAx4PZKlqfGcDQl1Y8I0WYlpZRfqdQZyRizyRizpYzd+gDbjTE7jTE5wMfAaHs5zwuBWfZ+72At71knOJqSIkN1NlWllH+pjkvVZsBel+dpdlpD4LgxJq9IukcicqeIpIhISnp6epUVtro4mpIi9V4GpZSfKfOsJCLzgSYeNj1mjPmiHHl4GpdpSkn3yBgzDZgGkJycXOJ+NYWj+ShSF+ZRSvmZMgODMWZoJfNIA1q4PG8O7AcOA/VEJMiuNTjS64RthzIA2Jl+2sclUUopd9XRlLQSaGuPQAoBrgdmG2uK0UXANfZ+Y4Dy1EBqhe12YPhhW81vFlNK1S6VHa56pYikAf2BOSIyz05vKiJzAezawD3APGATMNMYk2q/xMPAAyKyHavP4T+VKU9NFBGifQxKKf9SqbOSMeYz4DMP6fuBUS7P5wJzPey3E2vUUp3T/5yGLNt5hCt7lNjfrpRSPqGXqz7ywfi+/LAtnUHt4n1dFKWUcqOBwUcCAoTB7Rv5uhhKKVWM3nKrlFLKjQYGpZRSbjQwKKWUcqOBQSmllBsNDEoppdxoYFBKKeVGrJkpahYRSQd+rcAhcVhzM/mSlsH3+ftDGXydvz+Uwdf51+UytDLGlHnzVI0MDBUlIinGmBJXmNMy1I38/aEMvs7fH8rg6/y1DGXTpiSllFJuNDAopZRyU1cCwzRfFwAtgz/kD74vg6/zB9+Xwdf5g5ahVHWij0EppVT51ZUag1JKqXLSwKCUUspNrQkMIhLs6zI4iIj4KN9AX+bvUg6ffa98/d7tMgT6QRli7d8++VuISBP7t8/+HiLSWUTCfJj/QBFp46v8K6PGBwYR6SciHwP/FJEuPipDfxF5RUTGAphq7rixv4DvAI+LSIPqzt8uQx8RuQ/AGFPgg/z7isi/gYdFxCerH4lIsoi8B0z0xQlBRAJEJEZEvgJeger/W4hIDxFZADxr5++L72I3EfkReA5ryeDqzr+niHwLLARiqzt/b6jRgUFErgVeB74CwoAH7PRqu0oRkWuAV4GVwFARea46A5SInAO8BiwCWgHPisgl1ZW/XYb7sZZ4fVxERtpp1XLVLCKBIvI3rBEeS4GewJMi0rg68rfLECAirwJvAguABOApEYmorjKAMwicAoKBZiJynaN8VZ23WF4C3gXeMcbcUdV5luJxYJYx5kpjzD5H+ao6UxEJFpE3sb6Lr2Ctcz/Y3lajzrU1qrAetAW+NMa8D7wE1h+nmq9SOgP/M8a8B/wV6AtcKyL1qin/XsAmY8wM4C/AWuBSEWlRTfkDbAcuBe4CHgEwxuRXU4AOAPYA19qfwf1APyC8GvIGnCfkhcBFdhn+ARggr7rK4KID1jQLU4CbRCTaGFNQ1X8L+38uClhjjHkXQETaVOcJ0Q7QbYAMY8wUO22Y/b9YHc2socD3wPnGmK+AT4GOIhLki1p0ZdSowCAivxORB0Skv520BbhKRB4ClgFNgaki0rsay3AUCBWRWGPMb8BBoCXWyakq8u8nIu1cklYCzUWkhTHmGNZV83HgyqrIv4QyzAHW2b8zHE1K2P+MVZx/AfCRMWariIQaY/YDaVjz0FSZop+BMeZ/xpjjIjIMSMGqNbwgIh2rowwuJ7ztQA6wy/4ZIyItq+JiycP34C9AXxF5QkSWAv8EZohIL2/n7akM9sn3EHC+iFwiIp9jXay9Ajxo7+PVz6HIZ3DaGPOhMeaM/TwIyDfG5GmNoQrYzQUTgYftpH+LyOXA/4A/ARcAtxpjRgDpwNWOzq8qLsNwYAXQGHhLRGZinQwzAK92volIPRGZA3wH/E5EouxNWcCPwO/s51uAjUBDb3e8eShDpGOTMSbfGJMF/B9wu4jEGWO8esXs6TOw8z0OYIzJFpFooDWw35t5l1KGSDvd8Xc+BtxojBkGZGKdmL3arOWpDC4nvGTgpDEmFUgFngRet5s5vPL/XtJnYIw5CUwFrsaqOd4AHMD6f/Rqv08pZTgFvI3VxzHdGDMceAvoJyJeu1gr6W9gN6k5PufvgStFpL7WGKqAMSYfaA/8xRjzItaX/c9AO2PMAqyT4xZ79y+AbsDpKi7DU1hXSKew/glmAd8YY24AfgZG2sd56wolEqvN8l778QV2ejqwHOgqIn3scu4DBtonam/yWIYiX/rFdnnuBatTugrzP9/DPn2BVGPMfhGJEpG2XszfUxkcn4Gxf6cYY+ba+84FemAFiCovg20PEC0inwAPAauArcaYXC+enErM3xjzCjDEGPODMSYb+BwrWFXnZ/AVkAjUt5+nYNXks6s6f2MpsIPDbnufQV7Mt1r4bWAQkVtFZJBLW/1BoL7dXvcp1tXQ9XbNYAdwjb1fD6xAUdVlmAVsA64zxhw1xnxijJlu79ce6x/CW/nH2J1o04CZWO+vj4g0swPBcmAN8JJdk+gM7BEvdH6WUYa+ItLU3k/AGUCfwxoddALoWZlaUwXyD7IPqQfsFZFxWM1sSWebd0XL4EEvrCvmStecKlCG+kA88BvW/8JdQPvKNmlV5DOwmzQdemE17eVXJv9ylqGZnf86rKaje0QkDrgZ6AIcqeL8nf8LdhB21NizHOmVyb86+dWUGPYH1wT4EKvteAdWNP49cB9Wm90rdltuB+BjYBhWDeGPWH0MGcA9xpjN1VCG9lhfjBHGmAMichFWe+Yu4C5jzF4v5v8nY8xhe5+BWE1HKXant+PYF4HmWKOTbjXGbOEsVLAMK+3Of8fIi3OwqvI5wP3GmPXVlb+d/h5wE/AO8JJ9kqiwSnwGMVi1lhewTs5/McZsrYYyOL8LdjOeY3sUEGKMOVqNn0Eo0B+YjBUYq/0zsNMfwPo+tgX+bIzZWMX5u34GgfYAjPeAHcaYp87i7fuOMcYvfoBA+3c74H37cRDWUMz/YF0JzsOqskXY2/8L3G0/jgK6+qAMn2B9SQDaAFdWQf7/whr55Lrvn7GuzGOBaMfxjsc+KIPj82iE1ZRQnfnHAFF22vXANT76DMLstKHAaB+VIdLluxDgg/zD7bQBwBU++gyiXdKDffW/UNn8ffnj+wJYHxzfoN4AAAPHSURBVPQLwN+x2uIuwxoH7dgegDXSoDMwBuskfZ297QOgb00vQznyF6wrr0EuaVFYQxJXYDVxNfWDMjT3Uf4r7fwTfPgZOMrgD3+Hsy6Dfga+z98ffnzaxyAig7A6x+pjDbN7FsgFhjg6LY3VVvc08E9jzDvAt8CtIrIG6w9Y4aYKfypDOfM3wDNYHd4OlwB3A79g1ZTOehSOF8uQ5qP819r5Hzib/L1cBn/4O5xVGfQz8H3+fsOXUQlrVMktLs9fw+osGwusstMCsNr4ZgEt7LQmwDm1oQwVzH8mkGinjQYu8MFn4PUy+Dp/LYN/5O8PZfB1/v7y49vMIQLrbkFHW95NwN/sx2uBe+3HyVg3MdW6Mvg6f38og6/z1zL4R/7+UAZf5+8vPz5tSjLGZBpjso01xBGsEUbp9uNxWLeTfwV8BKyujWU4m/y9PezN12Xwdf5aBv/I3x/K4Ov8/YavI5MdfQOxqmdfA+faaedijQI6D2hW28vg6/z9oQy+zl/L4B/5+0MZfJ2/r3/85Qa3AqwZIQ8D3eyI/ARQYIz50dgzJNbyMvg6f38og6/z1zL4R/7+UAZf5+9bvo5MLhG6H9Yf40fg9rpYBl/n7w9l8HX+Wgb/yN8fyuDr/H354zd3PotIc+AW4EVjzbFS58rg6/z9oQy+zl/L4B/5+0MZfJ2/L/lNYFBKKeUf/KWPQSmllJ/QwKCUUsqNBgallFJuNDAopZRyo4FBKaWUGw0MSpVBRPJFZK2IpIrILyLygJSxfrKIJIrIjdVVRqW8SQODUmU7Y4xJMsZ0xpo7ZxTWuuOlSQQ0MKgaSe9jUKoMIpJhjIlyeX4O1qI0cVjLqL6HtdwjWMvK/iQiy4GOWMu8voO15OskYDDW7J1TjTFvVtubUKoCNDAoVYaigcFOOwZ0AE5hzZ+TJSJtsaZiThaRwcBfjTGX2vvfCTQyxjxnr4m8FLjWGLOrWt+MUuUQ5OsCKFVDOaZaDgZeFZEkIB9rfWBPLsaajO0a+3ks1iL1GhiU39HAoFQF2U1J+VjrgD+JtcZvd6w+u6ySDsNa5GVetRRSqUrQzmelKkBE4oE3gFeN1Q4bCxww1rrgt2DN4w9WE1O0y6HzgLtEJNh+nXYiEolSfkhrDEqVLVxE1mI1G+VhdTa/aG97DfhURK4FFgGn7fR1QJ6I/ALMAF7GGqm02l7xKx24orregFIVoZ3PSiml3GhTklJKKTcaGJRSSrnRwKCUUsqNBgallFJuNDAopZRyo4FBKaWUGw0MSiml3GhgUEop5eb/AeoUvJ5ksnwTAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the cumulative returns for neural network in train dataset\n",
    "data.strategy_returns_nn[:train_length+300].cumsum().plot()\n",
    "\n",
    "# Show the legend\n",
    "plt.legend()\n",
    "\n",
    "# Show the graph\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Plot cumulative returns in test dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the cumulative returns for neural network in test dataset\n",
    "data.strategy_returns_nn[train_length+300:].cumsum().plot()\n",
    "\n",
    "# Show the legend\n",
    "plt.legend()\n",
    "\n",
    "# Show the graph\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Strategy Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "|              |    Train |     Test |\n",
      "|--------------+----------+----------|\n",
      "| CAGR         | 7.25562  | 30.1549  |\n",
      "| Sharpe Ratio | 0.180996 |  2.06808 |\n"
     ]
    }
   ],
   "source": [
    "from tabulate import tabulate\n",
    "\n",
    "print (tabulate(\n",
    "        [             \n",
    "            ['CAGR', ret_train_nn, ret_test_nn],\n",
    "            ['Sharpe Ratio', sharpe_train_nn, sharpe_test_nn],\n",
    "        ], \n",
    "        headers=['Train', 'Test'], \n",
    "    tablefmt='orgtbl'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Accuracy of the neural network classifier\n",
    "\n",
    "The classification_report function from sklearn.metrics package is used to analyze the performance of the neural network. The classification_report takes as input the actual output and the predicted output."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "        0.0       0.43      0.78      0.56        88\n",
      "        1.0       0.65      0.28      0.39       126\n",
      "\n",
      "avg / total       0.56      0.49      0.46       214\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report\n",
    "print(classification_report(y_test_val, mlp.predict(X_test_val)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}