Spam Email Detection

A deep learning spam classifier using LSTM neural network trained on a balanced email dataset.

Results

• Test Accuracy: 94%
• Balanced dataset (equal spam/ham samples via downsampling)
• Early stopping to prevent overfitting

Tech Stack

Python, TensorFlow, Keras, NLTK, NumPy, Pandas, Matplotlib

How it Works

1. Preprocessing — removes punctuation, strips stop words, tokenises text
2. Balancing — downsamples ham emails to match spam count
3. Model — Embedding → LSTM → Dense → Sigmoid
4. Training — Adam optimizer, EarlyStopping + ReduceLROnPlateau callbacks

Model Architecture

Embedding → LSTM(16) → Dense(32, relu) → Dense(1, sigmoid)

Usage

predict("Congratulations! You won a free iPhone!")  # → 'spam'
predict("Hey, are we still meeting at 5pm?")        # → 'ham'

How to Run

1. Clone the repo
2. Open Detecting_Spam_Emails_Using_Tensorflow.ipynb in Colab or Jupyter
3. Run all cells (Runtime → Run all)