Fake News Detection using NLP in Jupyter Notebook

This project aims to detect fake news articles using Natural Language Processing (NLP) and Machine Learning techniques within a Jupyter Notebook environment. The model is trained to classify news as FAKE or REAL based on the article content.

Environment

Jupyter Notebook (Recommended)

Technologies Used

Pandas – data handling

NumPy – numerical operations

Matplotlib / Seaborn – data visualization

NLTK – text preprocessing

Scikit-learn – machine learning tools

Dataset

This project uses the Fake and Real News Dataset from Kaggle.

Click here to view/download the dataset on Kaggle

After downloading, extract the zip and place the Fake.csv file in your project directory.

The dataset used is Fake.csv, which contains:

title – Headline of the news article

text – Full content of the article

subject – Category or topic (e.g., News, Politics)

date – Date of publication

(Assumption: Dataset includes a label column with FAKE or REAL values)

How to Use (Jupyter Notebook)

Open your terminal or Anaconda Prompt and launch Jupyter Notebook:

bash Copy Edit jupyter notebook Open the notebook file: fake_news_detection.ipynb

Run each cell step-by-step:

Import libraries

Load dataset

Preprocess text

Vectorize text using TF-IDF

Train model

Evaluate results

Output

Accuracy Score of the model

Classification Report with precision, recall, F1-score

Confusion Matrix Heatmap for visual evaluation

Steps Followed in Notebook

Importing Libraries

Loading the Dataset

Text Cleaning – Lowercasing, removing punctuation and stopwords

TF-IDF Vectorization

Model Training – Logistic Regression

Evaluation – Accuracy, Confusion Matrix, Report

Future Scope

Experiment with other models (SVM, Naive Bayes, Random Forest)

Add LSTM or Transformer-based Deep Learning models

Build a web interface using Flask or Streamlit for live news testing

Folder Structure

python Copy Edit Fake-News-Detection/
│
├── fake_news_detection.ipynb
├── Fake.csv
├── README.md

Sample Output

Accuracy: 0.985 Classification Report: precision recall f1-score support

   Real       0.98      0.99      0.99       780
   Fake       0.99      0.98      0.98       776

accuracy                           0.98      1556

macro avg 0.98 0.98 0.98 1556 weighted avg 0.98 0.98 0.98 1556 Confusion Matrix:

Predicted Real Predicted Fake Real 770 10 Fake 14 762

Output Visualization

A heatmap showing the confusion matrix:

X-axis: Predicted labels

Y-axis: True labels

Annotated values