Email Spam Detection

A machine learning project for detecting spam emails using LSTM neural networks with TensorFlow and Keras.

Overview

This project implements a binary classification model to distinguish between spam and legitimate emails. The model uses natural language processing techniques and a Long Short-Term Memory (LSTM) architecture to achieve high accuracy in spam detection.

Features

• Data preprocessing with stopword removal and punctuation cleaning
• Balanced dataset handling to prevent bias
• LSTM-based deep learning model
• Comprehensive evaluation metrics including confusion matrix and classification report
• Model persistence for future predictions

Dataset

The project uses the spam_ham_dataset.csv containing labeled email samples. The dataset is balanced to ensure equal representation of spam and non-spam emails during training.

Model Architecture

• Embedding Layer: 32-dimensional word embeddings
• LSTM Layer: 16 units for sequence processing
• Dense Layer: 32 units with ReLU activation
• Output Layer: Single unit with sigmoid activation for binary classification

Results

• Test Accuracy: 96.33%
• Test Loss: 0.1537

The model demonstrates strong performance in distinguishing between spam and legitimate emails, with high precision and recall for both classes.

Requirements

• Python 3.12+
• TensorFlow 2.16.2
• NLTK
• WordCloud
• scikit-learn
• pandas
• numpy
• matplotlib
• seaborn

Usage

Open and run the Jupyter notebook in the notebook/ directory to:

1. Load and preprocess the dataset
2. Train the LSTM model
3. Evaluate performance metrics
4. Save the trained model

The trained model is saved in models/spam_detection_model.keras and can be loaded for future predictions.

Project Structure

Email-Spam-Detection-ML/
├── data/
│   └── spam_ham_dataset.csv
├── models/
│   └── spam_detection_model.keras
├── notebook/
│   └── notebook.ipynb
└── README.md