CM3070: University Final Project - Fake News Detection using Machine Learning and Deep learning techniques
Four notebooks used in the final project are available in a folder Notebooks. The detail of each notebook is as follows:
Notebook-1 : Used for data load, preprocessing, cleaning, data overview, feature extraction and training of traditional machine learning model (Logistic Regression, Multinomial Naive Bayes, and KNN).
Notebook-2 : Used for the training of LSTM (Multiple Runs performed to get the best parameters) and LSTM-CNN hybrid model.
Notebook-3 : LLM (BERT 1 billion parameters model) zero and few shot classifier testing using limited training.
Notebook-4 : LLM Fine-tuning (BERT), fine-tuning of the LLM is performed in this notebook. (Multiple runs are performed to get the best result using a separate notebook unlike LSTM keeping all runs within same notebook).
Models are saved in a folder Models