📧 Spam Detection using LSTM

📝 Project Overview

This project is a Spam Detection System built using Natural Language Processing (NLP) and Deep Learning. It classifies emails as Spam or Not Spam using an LSTM-based neural network trained on a balanced dataset.

⚙️ Features

• Preprocessing: Removes punctuation, stopwords, and subject lines.
• Tokenization & Padding: Converts text to sequences for deep learning.
• LSTM-based Model: Uses word embeddings and LSTM layers for classification.
• Binary Classification: Uses sigmoid activation for spam detection.
• Callbacks: Implements Early Stopping & ReduceLROnPlateau for better training.

---

📂 Dataset

• The dataset used is Spam Email Dataset, which contains email text and a binary label (spam: 1 for Spam, 0 for Not Spam).
• Balancing Strategy: Downsampling the majority class (ham emails) to avoid class imbalance.

---

🏗️ Model Architecture

model = tf.keras.models.Sequential()
model.add(tf.keras.layers.Embedding(input_dim=vocab_size, output_dim=32, input_length=max_length))
model.add(tf.keras.layers.LSTM(16))
model.add(tf.keras.layers.Dense(32, activation='relu'))
model.add(tf.keras.layers.Dense(1, activation='sigmoid'))

• Embedding Layer: Converts words into dense vector representations.
• LSTM Layer: Captures sequential dependencies in text.
• Dense Layer (ReLU): Extracts non-linear features.
• Output Layer (Sigmoid): Outputs probability for spam classification.

---

📊 Model Training

Train the Model

history = model.fit(
    train_padded, train_y,
    validation_data=(test_padded, test_y),
    epochs=20,
    callbacks=[early_stopping, reduce_lr]
)

Model Evaluation

Loss: 0.09856212884187698 Accuracy: 0.9799270033836365