Fake News Detection Project

Overview

• This project aims to detect fake news using machine learning techniques. It involves the classification of news articles into either real or fake categories based on their content.

Dataset

The dataset used consists of two main sources:

• True News: Contains genuine news articles.
• Fake News: Comprises fabricated or misleading articles.

Preprocessing

The dataset preprocessing steps involve:

• Labeling data: True news labeled as 0, fake news labeled as 1.
• Cleaning data: Text data was converted to lowercase, special characters removed, lemmatization applied, and stop words removed.

Dataset Analysis

• Checked for null values (no nulls found).
• Examined dataset balance: Balanced with similar proportions of true and fake news articles.

Model Training

Feature Extraction

• Used TF-IDF Vectorization: Converted text data into numerical features using TF-IDF vectorizer with 50,000 maximum features and considering unigrams and bigrams.

Model Selection

• Employed Multinomial Naive Bayes classifier for its effectiveness with text data.

Training and Evaluation

• Split dataset into train and test sets (80/20).
• Trained the model on the training set and evaluated performance on the test set.
• Obtained classification reports and accuracy scores for both test and train sets.

Results

• Achieved accuracy scores on both train and test sets.
• Classification reports showcase precision, recall, and F1-score for each class.

Conclusion

The Multinomial Naive Bayes model demonstrated reasonable accuracy in detecting fake news based on the provided dataset and features.

Instructions to Run

1. Ensure necessary Python libraries are installed (e.g., pandas, numpy, scikit-learn, nltk).
2. Run the provided Python script after adjusting the file paths if needed.