FakeNews

Dataset(train.csv, test.csv) from here: https://www.kaggle.com/jsvishnuj/fakenews-detection-using-lstm-neural-network/data

I have a main folder called FakeNews with all the .py files. Inside FakeNews, there is a folder named Kaggle where train.csv, test.csv are stored. When you run DataPreprocess.py two files are generated inside the folder: train_data.csv, test_data.csv.

We use a GloVe embeddings file to train the network: https://www.kaggle.com/terenceliu4444/glove6b100dtxt

(optional)The pretrained model.h5 can also be downloaded from here: https://drive.google.com/open?id=1nx6tFo4o2ZhHF6qSds3wwnHJg6UvUv7T To use it in the Django app, place it in env/src/fakenews/static

To install and run the django app for Windows: Make sure you have pip and virtualenv installed.

1. Create a virtual environment: virtualenv env
2. To activate it run : env/Scripts: activate
3. To get the required packages place requirements.txt inside env and run: pip install -r requirements.txt
4. Create a source folder and go: mkdir src; cd src
5. Start project: django-admin startproject fakenews
6. Download the project directory (fakenews) from here and put it inside the src folder
7. To run the Django server: env\src\fakenews python manage.py runserver
8. Open a broswer and go to: http://127.0.0.1:8000/
9. You can now use the app!

To uninstall it just delete the env folder.

To use only the neural network:

1. Download DataPreprocess.py, model.py
2. Download the dataset. You should have two files inside Kaggle folder like this: Kaggle/train.csv and Kaggle/test.csv

To train the network from scratch and generate model.h5 run --python model.py and type (1) in the terminal when asked. To make predictions follow the same procedure but type (2). An rf_pred.csv file will be created inside the folder "Kaggle".

There is also a back-up app which we built just in case. To use the backup app:

1. Download tkinter_GUI.py and run python tkinter_GUI.py! You have to insert title, author and text to get a prediction.