fake_news_classifier

Project for the UE15CS322 Data Analytics Course

Team members

Hiranmaya Gundu 01FB15ECS127
Rahul Mayuranath 01FB15ECS225
Ravi Shreyas Anupindi 01FB15ECS234

Credits

Thanks to https://github.com/genyunus/Detecting_Fake_News/ for the cross_validation script.

Words of Advice:

The models take a fair amount of time to train and test, especially Random Forests and LSTM. This is probably owing to the K-folds cross validation taking quite some time.
Additionally, to run the models against the entire dataset(30,000 articles) you will require at least 16GB of RAM. To work around this, navigate to the file start.py and make changes to the line:

df = df.sample(1000)

This line can be commented out, or you can manually vary the size of the dataset. For LSTM, navigate to the file RNN.py and make changes to the line:

df = df.sample(10000)

Dataset

Link to the dataset: https://drive.google.com/open?id=1Q3ZO3trNfBg0uSlkz73V6XfSRNx7hO7k Create a folder called data inside the main directory and add the csv file there.

Getting started

These instructions will help you in setting up the python environment for running the models and validating results. After cloning the repository, create a python3 virtual environment to install the required packages.

cd fake_news_classifier
python3 -m venv <environment_name>  
source <environment_name>/bin/activate

This creates the virtual_environment for the project and sets the python interpreter to the project interpreter. Installing the required packages:

pip install -r requirements.txt

The package also requires additional nltk packages, including the corpus of stop words. These can be downloaded by running the command

import nltk
nltk.download()

from inside the python shell.
Generating the word clouds:

python wcloud.py

Running the models:

python main.py