SMS Spam Detection 📱✉️

This project focuses on building a machine learning system that classifies SMS messages into two categories:

• Ham (legitimate messages)
• Spam (undisered messages)

---

📂 Data Understanding & Preprocessing

• The dataset contains thousands of SMS messages labeled as Ham or Spam.
• The distribution is imbalanced (Ham >> Spam)( [Ham / Spam ] = 6, which means that the ham messages are 6 times more than spam ones.
• To make the data usable for machine learning, the text was cleaned and transformed into numerical features using TF-IDF Vectorization.
• Words were lowercased, special characters were removed, and stopwords were filtered out.

---

⚙️ Machine Learning Pipeline

To ensure reproducibility and a clean workflow, a Pipeline was used that combines:

1. TF-IDF Vectorizer → transforms raw text into numerical features.
2. Model (Naive Bayes or XGBoost) → learns to classify the messages.

This way, preprocessing and model training are linked together and can be reused consistently.

---

🔍 Model Selection with GridSearchCV

• Hyperparameter tuning was done using GridSearchCV.
• For each model (Naive Bayes and XGBoost), a grid of parameters was tested.
• GridSearch ensures that the best parameters are selected based on cross-validation performance, avoiding overfitting.

---

📊 Results & Interpretation

Naive Bayes (NB)

• Precision (Spam): 1.0 ✅
• Recall (Spam): slightly lower than XGBoost (27 Spam messages were missed).
• Confusion Matrix:
  • No Ham misclassified as Spam (0 false positives).
  • Safe model, user never loses important messages.

XGBoost (XGB)

• Precision (Spam): slightly lower than 1.0 (introduced 8 false positives).
• Recall (Spam): higher than NB (detected more Spam, only 23 missed).
• Confusion Matrix:
  • Caught more Spam, but at the cost of flagging some Ham messages as Spam.

---

✅ Final Choice

For this project, Naive Bayes was chosen because:

• It achieved perfect precision (1.0) on Spam.
• Even though XGBoost had better recall, NB ensures no false alarms.
• In practice, it is more important not to lose legitimate messages than to catch every single Spam.

---

🚀 How to Run

1. Clone the repository:

   git clone https://github.com/yourusername/sms-spam-detector.git
   cd sms-spam-detector

2. Create and activate a virtual environment:

   python3 -m venv venv
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Open Jupyter Notebook:

   jupyter notebook spam-detection.ipynb

5. Run the UI interface:

   streamlit run app.py

6. Try your own SMS and see the results.