SMS Spam Detector

A machine learning model that classifies SMS messages as spam or not spam, deployed as an interactive web app using Streamlit.

---

Dataset

• Source: SMS Spam Collection Dataset — Kaggle
• 5,572 labeled SMS messages (ham / spam)

---

Pipeline

1. Preprocessing

• Lowercasing, punctuation and stopword removal
• Tokenization using nltk
• Stemming with PorterStemmer

2. Feature Extraction

• Text vectorized using TF-IDF (TfidfVectorizer)

3. Model

• Random Forest Classifier
  • n_estimators = 200
  • random_state = 2
  • Tuned splitting criteria for better generalization

---

Results

Metric	Score
Accuracy	~97%
Precision	100%
False Positives	0

Confusion Matrix:

	Predicted: Not Spam	Predicted: Spam
Actually Not Spam	896	0
Actually Spam	29	109

Test set performance. Real-world results may vary due to class imbalance and evolving spam patterns.

---

Tech Stack

Tool	Purpose
pandas, numpy	Data manipulation
nltk	NLP preprocessing
scikit-learn	Modeling & evaluation
matplotlib, seaborn	EDA & visualization
streamlit	Web app deployment
pickle	Model serialization

---

Run Locally

git clone https://github.com/maheen8q/spam-detector.git
cd spam-detector
pip install -r requirements.txt
streamlit run app.py

---

📁 Project Structure

spam-detector/
│
├── spam-detection.ipynb   # EDA, preprocessing, modeling
├── app.py                 # Streamlit app
├── model.pkl              # Trained Random Forest model
├── vectorizer.pkl         # Fitted TF-IDF vectorizer
└── README.md

---

https://spam-detector-yvxs9mmkn6sumxug2ch