📧 Email Spam Classifier using LSTM and NLP

A deep learning-based spam detection system that uses Natural Language Processing (NLP) techniques and a Long Short-Term Memory (LSTM) neural network to classify emails as spam or not spam. This project includes data visualization, preprocessing, model training, and a prediction interface for real-time email classification.

---

📌 Features

• 📊 Exploratory Data Analysis with WordClouds and Seaborn plots
• 🔁 Balanced dataset using downsampling for fair training
• 🧹 Text preprocessing (punctuation removal, stopword filtering)
• 🧠 LSTM-based deep learning model using Keras and TensorFlow
• ✅ Real-time email spam prediction from user input
• 💾 Model and tokenizer saved for future predictions

---

🧰 Libraries Used

• numpy, pandas, matplotlib, seaborn
• nltk, wordcloud
• scikit-learn
• tensorflow, keras
• pickle

---

📂 Dataset

The project uses an emails.csv dataset with two columns:

• text: Raw email content
• spam: Binary label (0 = Not Spam, 1 = Spam)

---

🧪 Model Architecture

• Text tokenization and padding
• Embedding Layer to learn word representations
• LSTM Layer with 16 units to capture sequence dependencies
• Dense layers with ReLU and Sigmoid activations
• Loss: BinaryCrossentropy
• Optimizer: Adam
• EarlyStopping and ReduceLROnPlateau callbacks for better training control

---

📈 Performance

• Training and validation accuracy visualized across epochs
• Final test loss and accuracy printed after evaluation
• Real-time predictions based on new email input from the user

---

💾 Output Artifacts

• Trained model saved as: spam_detector_model.h5
• Tokenizer saved as: tokenizer.pickle
• You can reuse these for deployment or further predictions.

---

🧠 Sample Prediction

Enter the email text to check if it is spam: 
Congratulations! You've won a free iPhone! Click here to claim now.

The email is classified as: Spam