Fake News Detection Using Machine Learning

This project is focused on detecting fake news using machine learning techniques. The dataset is preprocessed and analyzed before training a classification model. The code explores various steps, from shuffling the dataset and removing irrelevant text to visualizing word frequency and training different classifiers.

Overview

This project includes:

• Shuffling the dataset to avoid biases
• Text preprocessing to remove stopwords and punctuation
• Word cloud and frequency bar chart visualizations
• Machine learning models for classification (Logistic Regression and Decision Tree)

Dataset

https://www.kaggle.com/datasets/subho117/fake-news-detection-using-machine-learning

Setup

1. Clone the repository:

   git clone https://github.com/dharmik2101/Fake-News-Detection-Using-Machine-Learning.git
   cd fake-news-detection

2. Install dependencies:

   pip install -r requirements.txt

3. Download NLTK resources: Run the script to download stopwords and tokenizer data:

   import nltk
   nltk.download('punkt')
   nltk.download('stopwords')

Data Preprocessing

• Shuffle and Reset Index: Randomizes the dataset to prevent order bias and removes unnecessary index column.
• Text Preprocessing: Cleans the text by removing stopwords and punctuation using NLTK. The cleaned text is then ready for model input.

Exploratory Data Analysis (EDA)

• Word Cloud Visualization: Generates separate word clouds for real and fake news to observe common terms.
• Frequency Bar Chart: Shows the 20 most common words, aiding insight into word patterns across classes.

Model Training and Evaluation

1. Logistic Regression: Basic linear classifier, with accuracy evaluated on training and test sets.
2. Decision Tree Classifier: Uses a decision tree model, often achieving higher accuracy.

Model Accuracy:

• Logistic Regression: Training ~99.4%, Test ~98.9%
• Decision Tree: Training ~99.9%, Test ~99.5%

3. Confusion Matrix: Visualizes the Decision Tree model's performance in predicting real vs. fake news.

Results

• Both classifiers demonstrate high accuracy, with Decision Tree outperforming Logistic Regression.
• The model successfully detects fake news with high precision.

Reference

Geeks for Geeks.

---