📰 Fake News Detection App

An NLP-based Machine Learning project that classifies news articles as Fake or Real using TF-IDF and Logistic Regression, with model explainability using top influencing words.

---

🚀 Live Demo

The application is deployed on Streamlit Cloud.

🔗 Click here to try the Live App

You can test the model by entering any news article text and instantly receive:

• Prediction (Fake / Real)
• Confidence Score
• Top Influential Words
• Model Explanation

---

📌 Project Overview

This project builds a binary text classification system that detects whether a news article is Fake or Real.

The model is trained using TF-IDF vectorization and Logistic Regression, achieving 98.75% accuracy on the test dataset.

The app also provides:

• ✅ Prediction result
• 📊 Confidence score
• 🔍 Top influencing words
• 📖 Explanation of classification

---

🛠️ Tech Stack

• Python
• Scikit-learn
• Pandas
• NLTK
• Streamlit
• Pickle (Model Saving)

---

📂 Dataset

Dataset used:

• Fake.csv
• True.csv

The dataset contains labeled real and fake news articles.

⚠️ Note:
The dataset is not included in this repository due to GitHub file size limitations. It is used only for training the model.

---

⚙️ Project Workflow

1️⃣ Data Collection

• Collected labeled dataset from Kaggle.

2️⃣ Text Preprocessing

• Lowercasing
• Removing punctuation
• Removing stopwords (NLTK)
• Stemming

3️⃣ Feature Extraction

• TF-IDF Vectorization

4️⃣ Model Training

• Logistic Regression classifier

5️⃣ Model Evaluation

• Accuracy: 98.75%
• Precision, Recall, F1-score evaluated using classification report

6️⃣ Model Saving

• fake_news_model.pkl
• tfidf_vectorizer.pkl

7️⃣ Streamlit App

• User enters news text
• Model predicts Fake or Real
• Displays:
  • Confidence score
  • Top influencing words
  • Explanation message

---

🧠 Explainability Feature

The app displays the top TF-IDF weighted words influencing the prediction.

Example:

reuter (impact score: 2.18)
said (impact score: 1.18)
monday (impact score: 0.59)

This makes the model more transparent and interpretable.

---

📊 Model Performance

Metric	Score
Accuracy	98.75%
Precision	0.99
Recall	0.99
F1-score	0.99

---

⚠️ Important Note

This model learns dataset-specific linguistic patterns (e.g., Reuters-style formatting vs clickbait tone).

It is not a universal fact-checking system, but a pattern-based text classification model.

---

📁 Project Structure

news-article-classification/
│
├── app.py
├── fake_news_model.pkl
├── tfidf_vectorizer.pkl
├── News_Article_Classification.ipynb
└── README.md

---

▶️ How to Run Locally

1️⃣ Clone Repository

git clone https://github.com/YOUR_USERNAME/news-article-classification.git
cd news-article-classification

2️⃣ Install Dependencies

pip install -r requirements.txt

3️⃣ Run Streamlit App

streamlit run app.py

App will open at:

http://localhost:8501

---

📌 Future Improvements

• Add deep learning model (LSTM / BERT)
• Add model comparison
• Improve UI styling
• Deploy on cloud permanently

---

👨‍💻 Author

Vidit Kumar

AI & Machine Learning Enthusiast
Passionate about NLP, Web Development, and AI Systems

---

⭐ If you found this project useful, feel free to star the repository!

---