Major update

pyeconomics/__init__.py

@@ -1,18 +1,22 @@
 # pyeconomics/__init__.py
 
+# AI imports
+from pyeconomics.ai import ai_regression_analysis
+
 # API imports
 from pyeconomics.api import (
-    bitcoin_s2f_data,
-    load_bitcoin_data,
     FredClient,
+    bitcoin_s2f_data,
     fred_client,
+    load_bitcoin_data,
     load_from_cache,
     save_to_cache
 )
 
 # Data imports
 from pyeconomics.data import (
     BalancedApproachRuleParameters,
+    ChatGPTParameters,
     EconomicIndicators,
     FirstDifferenceRuleParameters,
     TaylorRuleParameters,
@@ -55,7 +59,9 @@
 from pyeconomics.models.stock_to_flow import (
     bitcoin_s2f_forecast,
     calculate_model_values,
-    fit_model,
+    fit_regression_model,
+    plot_s2f_model,
+    plot_s2f_prediction_model,
     power_law_function
 )
 
@@ -73,10 +79,12 @@
     fetch_historical_fed_funds_rate,
     print_fred_series_names
 )
+from pyeconomics.utils.utils import get_forecast_tickers
 
 # Exported symbols
 __all__ = [
     'BalancedApproachRuleParameters',
+    'ChatGPTParameters',
     'EconomicIndicators',
     'FirstDifferenceRuleParameters',
     'TaylorRuleParameters',
@@ -94,15 +102,18 @@
     'historical_balanced_approach_rule',
     'historical_first_difference_rule',
     'historical_taylor_rule',
+    'ai_regression_analysis',
     'plot_historical_rule_estimates',
     'plot_historical_bar_basr_rule',
     'plot_historical_fdr',
     'plot_historical_taylor_rule',
+    'plot_s2f_model',
+    'plot_s2f_prediction_model',
     'verbose_monetary_policy_rules',
     'taylor_rule',
     'bitcoin_s2f_forecast',
     'calculate_model_values',
-    'fit_model',
+    'fit_regression_model',
     'power_law_function',
     'verbose_balanced_approach_rule',
     'verbose_first_difference_rule',
@@ -121,5 +132,6 @@
     'variance_inflation_factor_test',
     'verbose_model_diagnostics',
     'fetch_historical_fed_funds_rate',
-    'print_fred_series_names'
+    'print_fred_series_names',
+    'get_forecast_tickers'
 ]

pyeconomics/ai/__init__.py

@@ -0,0 +1,7 @@
+# pyeconomics/ai/__init__.py
+
+from .ai_regression_analysis import ai_regression_analysis
+
+__all__ = [
+    'ai_regression_analysis'
+]

pyeconomics/ai/ai_regression_analysis.py

@@ -0,0 +1,120 @@
+# pyeconomics/ai/ai_regression_analysis.py
+
+import os
+import openai
+from statsmodels.regression.linear_model import RegressionResults
+from statsmodels.stats.diagnostic import het_breuschpagan, linear_reset
+from statsmodels.stats.stattools import durbin_watson, jarque_bera
+from statsmodels.stats.stattools import omni_normtest
+from typing import Union
+from IPython.display import display, Markdown
+
+from pyeconomics.api.openai_api import load_prompt, initialize_openai_client
+from ..data.model_parameters import ChatGPTParameters
+
+
+def ai_regression_analysis(
+    model: RegressionResults,
+    ai_params: ChatGPTParameters = ChatGPTParameters()
+) -> Union[None, str]:
+    """
+    Generate an AI-based analysis of the OLS regression results.
+
+    Args:
+        model (RegressionResults): A statsmodels OLS Regression Results object.
+        ai_params (ChatGPTParameters): Instance containing the AI parameters.
+
+    Returns:
+        Union[None, str]: AI-generated analysis paragraph or None if Markdown
+            output is enabled.
+    """
+    # Ensure the OpenAI client is initialized only when this function is called
+    initialize_openai_client()
+
+    # Extract the needed estimates from the ai_model
+    slope_name = model.params.index[1]
+    estimates = {
+        'R_squared': model.rsquared,
+        'Adj_R_squared': model.rsquared_adj,
+        'F_statistic': model.fvalue,
+        'Prob_F_statistic': model.f_pvalue,
+        't_statistic': model.tvalues[slope_name],
+        'P_value': model.pvalues[slope_name],
+        'Omnibus': omni_normtest(model.resid)[0],
+        'Prob(Omnibus)': omni_normtest(model.resid)[1],
+        'Jarque-Bera': jarque_bera(model.resid)[0],
+        'Prob(JB)': jarque_bera(model.resid)[1],
+        'Skewness': model.resid.skew(),
+        'Kurtosis': model.resid.kurtosis() + 3,
+        'Breusch_Pagan': het_breuschpagan(model.resid, model.model.exog)[0],
+        'Prob(BP)': het_breuschpagan(model.resid, model.model.exog)[1],
+        'Durbin_Watson': durbin_watson(model.resid),
+        'Condition_Number': model.condition_number,
+        'Intercept': model.params['const'],
+        'Slope': model.params[slope_name],
+        'Slope_Name': slope_name,
+        'Ramsey_Reset': linear_reset(model, power=2, use_f=True).fvalue,
+        'Prob(RR)': linear_reset(model, power=2, use_f=True).pvalue
+    }
+
+    # Load the prompt template
+    prompt_filepath = 'ai_regression_analysis.txt'
+    prompt_file_path = os.path.join(
+        os.path.dirname(__file__),
+        'prompts',
+        prompt_filepath
+    )
+    prompt_template = load_prompt(prompt_file_path)
+
+    # Format the prompt with data
+    prompt = prompt_template.format(
+        r_squared=round(estimates['R_squared'], 3),
+        adj_r_squared=round(estimates['Adj_R_squared'], 3),
+        f_statistic=round(estimates['F_statistic'], 2),
+        prob_f_statistic=round(estimates['Prob_F_statistic'], 2),
+        t_statistic=round(estimates['t_statistic'], 3),
+        p_value=round(estimates['P_value'], 3),
+        omnibus=round(estimates['Omnibus'], 3),
+        prob_omnibus=round(estimates['Prob(Omnibus)'], 3),
+        jarque_bera=round(estimates['Jarque-Bera'], 3),
+        prob_jb=round(estimates['Prob(JB)'], 3),
+        skewness=round(estimates['Skewness'], 3),
+        kurtosis=round(estimates['Kurtosis'], 3),
+        breusch_pagan=round(estimates['Breusch_Pagan'], 3),
+        prob_breusch_pagan=round(estimates['Breusch_Pagan'], 3),
+        ramsey_reset=round(estimates['Ramsey_Reset'], 3),
+        prob_ramsey_reset=round(estimates['Prob(RR)'], 3),
+        durbin_watson=round(estimates['Durbin_Watson'], 3),
+        condition_number=round(estimates['Condition_Number'], 2),
+        intercept=round(estimates['Intercept'], 4),
+        slope=round(estimates['Slope'], 4),
+        slope_name=estimates['Slope_Name'],
+        ai_model=ai_params.ai_model,
+    )
+
+    # Generate the AI analysis
+    response = openai.chat.completions.create(
+        model=ai_params.ai_model,
+        messages=[
+            {"role": "system",
+             "content": "You are an expert data scientist. Your task is to "
+                        "analyze the OLS regression results and provide a "
+                        "detailed interpretation in markdown format for a "
+                        "Jupyter notebook."
+             },
+            {"role": "user", "content": prompt}
+        ],
+        max_tokens=ai_params.max_tokens
+    )
+
+    # Remove the ```markdown``` delimiters from the response content
+    analysis = response.choices[0].message.content.strip()
+    if analysis.startswith("```markdown"):
+        analysis = analysis[len("```markdown"):].strip()
+    if analysis.endswith("```"):
+        analysis = analysis[:-len("```")].strip()
+
+    if ai_params.markdown:
+        display(Markdown(analysis))
+    else:
+        return analysis

pyeconomics/ai/balanced_approach_rule.py

@@ -4,7 +4,7 @@
 
 import openai
 
-from pyeconomics.api.openai_api import load_prompt
+from pyeconomics.api.openai_api import load_prompt, initialize_openai_client
 from pyeconomics.utils.utils import encode_image
 
 
@@ -23,14 +23,17 @@ def balanced_approach_rule(
         max_tokens (int): Maximum number of tokens for the AI response.
             Defaults to 500 which may cost a few cents per call. Adjust as
             needed. See https://openai.com/api/pricing/ for details.
-        model (str): The OpenAI model to use for the analysis. Defaults to
+        model (str): The OpenAI ai_model to use for the analysis. Defaults to
             'gpt-4o'. Other models are available, such as 'gpt-4-turbo' and
             'gpt-3.5-turbo'. See https://platform.openai.com/docs/models for
             more information.
 
     Returns:
         str: AI-generated analysis paragraph.
     """
+    # Ensure the OpenAI client is initialized only when this function is called
+    initialize_openai_client()
+
     # Construct the full path to the prompt file
     prompt_file_path = os.path.join(
         os.path.dirname(__file__),
@@ -89,7 +92,7 @@ def plot_interpretation(
         max_tokens (int): Maximum number of tokens for the AI response. Defaults
             to 500 which may cost a few cents per call. Adjust as needed. See
             https://openai.com/api/pricing/ for details.
-        model (str): The OpenAI model to use for the analysis. Defaults to
+        model (str): The OpenAI ai_model to use for the analysis. Defaults to
             'gpt-4o'. Other models are available, such as 'gpt-4-turbo' and
             'gpt-3.5-turbo'. See https://platform.openai.com/docs/models for
             more information.

pyeconomics/ai/first_difference_rule.py

@@ -4,7 +4,7 @@
 
 import openai
 
-from pyeconomics.api.openai_api import load_prompt
+from pyeconomics.api.openai_api import load_prompt, initialize_openai_client
 from pyeconomics.utils.utils import encode_image
 
 
@@ -21,14 +21,17 @@ def first_difference_rule(
         max_tokens (int): Maximum number of tokens for the AI response. Defaults
             to 500 which may cost a few cents per call. Adjust as needed. See
             https://openai.com/api/pricing/ for details.
-        model (str): The OpenAI model to use for the analysis. Defaults to
+        model (str): The OpenAI ai_model to use for the analysis. Defaults to
             'gpt-4o'. Other models are available, such as 'gpt-4-turbo' and
             'gpt-3.5-turbo'. See https://platform.openai.com/docs/models for
             more information.
 
     Returns:
         str: AI-generated analysis paragraph.
     """
+    # Ensure the OpenAI client is initialized only when this function is called
+    initialize_openai_client()
+
     # Construct the full path to the prompt file
     prompt_file_path = os.path.join(
         os.path.dirname(__file__),
@@ -89,7 +92,7 @@ def plot_interpretation(
         max_tokens (int): Maximum number of tokens for the AI response. Defaults
             to 500 which may cost a few cents per call. Adjust as needed. See
             https://openai.com/api/pricing/ for details.
-        model (str): The OpenAI model to use for the analysis. Defaults to
+        model (str): The OpenAI ai_model to use for the analysis. Defaults to
             'gpt-4o'. Other models are available, such as 'gpt-4-turbo' and
             'gpt-3.5-turbo'. See https://platform.openai.com/docs/models for
             more information.

pyeconomics/ai/monetary_policy_rules.py

@@ -4,7 +4,7 @@
 
 import openai
 
-from pyeconomics.api.openai_api import load_prompt
+from pyeconomics.api.openai_api import load_prompt, initialize_openai_client
 from pyeconomics.data.economic_indicators import EconomicIndicators
 from pyeconomics.data.model_parameters import MonetaryPolicyRulesParameters
 
@@ -29,6 +29,9 @@ def monetary_policy_rules(
     Returns:
         str: AI-generated analysis paragraph.
     """
+    # Ensure the OpenAI client is initialized only when this function is called
+    initialize_openai_client()
+
     if params.rho > 0.0 or params.apply_elb:
         prompt_filepath = 'adjusted_monetary_policy_rules.txt'
     else:
@@ -46,25 +49,42 @@ def monetary_policy_rules(
 
     # Format the prompt with data
     prompt = prompt_template.format(
-        taylor_rule=round(estimates['Taylor Rule (TR)'], 2),
+        taylor_rule=round(
+            estimates['Taylor Rule (TR)'], 2),
         balanced_approach_rule=round(
             estimates['Balanced Approach Rule (BAR)'], 2),
         balanced_approach_shortfalls_rule=round(
             estimates['Balanced Approach Shortfalls Rule (BASR)'], 2),
         first_difference_rule=round(
             estimates['First Difference Rule (FDR)'], 2),
-        current_fed_rate=round(indicators.current_fed_rate, 2),
+        current_fed_rate=round(
+            indicators.current_fed_rate, 2),
+        current_inflation_rate=round(
+            indicators.current_inflation_rate, 2),
+        inflation_target=round(
+            params.inflation_target, 2),
+        current_unemployment_rate=round(
+            indicators.current_unemployment_rate, 2),
+        natural_unemployment_rate=round(
+            indicators.natural_unemployment_rate, 2),
+        long_term_real_interest_rate=round(
+            indicators.long_term_real_interest_rate, 2),
+        inflation_gap=round(
+            indicators.current_inflation_rate - params.inflation_target, 2),
+        unemployment_gap=round(
+            indicators.natural_unemployment_rate -
+            indicators.current_unemployment_rate, 2),
         rho=params.rho
     )
 
     response = openai.chat.completions.create(
         model=params.model,
         messages=[
             {"role": "system",
-             "content": "You are the Federal Open Market Committee (FOMC) of "
-                        "the Federal Reserve System. Your task is to analyze "
-                        "monetary policy rule estimates and provide a policy "
-                        "prescription based on the given data."
+             "content": "Act as the Federal Open Market Committee (FOMC) of "
+                        "the Federal Reserve System (the Fed) that is charged "
+                        "with making key decisions about interest rates and "
+                        "the growth of the United States money supply."
              },
             {"role": "user", "content": prompt}
         ],