Spam / Ham Email Classification

📌 Overview

This project is a Spam/Ham email classification system built using Jupyter Notebook.
It trains a machine learning pipeline to classify emails/messages as spam or ham using
Count/TF-IDF Vectorizer and Multinomial Naive Bayes (MNB).

The project also includes a simple Streamlit interface for interactive predictions.

---

⚙️ Technology Used

• Python (Jupyter Notebook)
• pandas, numpy → data loading & preprocessing
• scikit-learn → CountVectorizer / TfidfVectorizer, MultinomialNB, Pipeline, evaluation
• joblib → saving/loading trained model
• nltk → optional stopword handling
• matplotlib, seaborn → plots & confusion matrix
• Streamlit → simple demo app

---

🔄 Process

1. Dataset → Load spam_ham_database.csv (label + text columns).
2. Preprocessing → lowercase, remove URLs/emails, normalize numbers, clean special characters (lemmatization optional).
3. Train/Test Split → 80% training, 20% testing with stratification to keep class balance.
4. Vectorization → Convert text into numeric features using CountVectorizer / TF-IDF with unigrams + bigrams.
5. Model Training → Multinomial Naive Bayes classifier trained on the vectorized data.
6. Evaluation → Accuracy, Precision, Recall, F1-score, and Confusion Matrix.
7. Artifacts → Save the trained pipeline with joblib for reuse.
8. Streamlit Demo → Input text → cleaned → classified as Spam/Ham with probability.

---

✅ Outcome

• Achieved high accuracy (~90–95% depending on preprocessing).
• Built a reproducible ML pipeline (preprocessing + vectorizer + model).
• Demonstrated results interactively with Streamlit.
• Clear separation of stages: preprocessing → training → evaluation → serving.

---

🖥️ Streamlit Interface

The Streamlit app provides a simple way to test the model.

Run locally:

# Install dependencies
pip install -r requirements.txt

# Train & save pipeline (run the notebook first if needed)
jupyter notebook notebook.ipynb

# Run Streamlit demo
streamlit run streamlit_app.py