Harshal Bansal

Project Title: Fake News Detection Using Hybrid CNN-LSTM Model

Description:
Developed an end-to-end fake news detection system leveraging Natural Language Processing and deep learning techniques. Utilized a hybrid CNN-LSTM architecture to capture both spatial and temporal dependencies in news articles. Preprocessed textual data using tokenization, padding, and embedding with GloVe vectors.

Dataset: Used the Fake and Real News Dataset from Kaggle, comprising ~44,000 news articles (23,000 real, 21,000 fake).

Model Architecture:

CNN layers to extract local features and patterns (e.g., phrases and keyword combinations)

LSTM layers to capture sequence dependencies across the articles

Embedding Layer: Pre-trained GloVe (100d)

Dropout & Batch Normalization for regularization

Results:

Achieved accuracy of 94.6% on the test set

Precision: 95.1%, Recall: 93.8%, F1-score: 94.4%

Trained over 5 epochs with early stopping and model checkpointing

Tools/Tech: Python, TensorFlow/Keras, NumPy, Pandas, Matplotlib, NLTK, Scikit-learn

Contribution: Built the model from scratch, handled preprocessing, tuning, and evaluation