Detecting Fake news with Recurrent Neural Networks
Updated: Nov 29, 2020
Today's world unfortunately, is full of misinformation and online trickery spreading fake news to influence perceptions and beliefs. We can investigate its authenticity before spreading the news by sharing with others.
One way to do it is to trace it to the source and see if many other media platforms are mentioning the news in credible websites. But if you are part of the media company itself ( and you don't have a correspondent on ground to verify ) - how do you validate its authenticity ? Or what if there was an AI based tool, which gives a high accuracy of predicting whether a piece of news is real or fake ? Read on to find out how ...
I have used a database of around 50,000 records of real and fake news to train a recurrent neural network model.
Perform exploratory analysis of the training data - remove stop word which add no information and training value.
Prepare the data by performing tokenising and padding.
Train using Bi-Directional RNN and LSTM
Make prediction : # if the predicted value is >0.5 it is real else it is fake
Get accuracy score : Model Accuracy : 0.9968819599109131