Spam Detector App

A Machine Learning project to detect spam messages using Natural Language Processing (NLP), TF-IDF vectorization, SMOTE for imbalance handling, and a Logistic Regression classifier — all wrapped up in Streamlit web app.

---

Model Accuracy: ~99% on test set

---

Project Structure

project-root/
│
├── data/
│   ├── spam.csv                # Original dataset
│   ├── finalmodel.pkl          # Trained ML model
│   ├── vectorizer.pkl          # Saved TF-IDF vectorizer
│   ├── feature.pkl           
│   └── label.pkl         
│
├── notebook/
│   └── eda.ipynb               # Exploratory Data Analysis
│
├── src/
│   ├── preprocessing.ipynb     # Preprocessing pipeline
│   └── training.ipynb          # Model training, tuning, evaluation
│
├── app.py                      # Streamlit app
└── README.md                   # You are here!

---

Features

• Cleans & lemmatizes text
• TF-IDF vectorization
• Text length feature
• SMOTE for handling imbalanced classes
• Classifies using Logistic Regression
• Also tested with Multinomial Naive Bayes
• Built with reusability using joblib
• Streamlit app for user interaction

---

Tech Stack

• Python
• Pandas, NumPy, Matplotlib, Seaborn
• scikit-learn, imblearn, nltk
• Streamlit
• joblib

---

Performance

Model	Accuracy	Precision	Recall	F1-Score
Logistic Regression	99%	0.99	0.99	0.99
MultinomialNB	96%	0.93	1.00	0.96

---

How to Run

1. Clone the repo

   git clone https://github.com/sarfraspc/spam-detector.git

2. Install requirements

   pip install -r requirements.txt

3. Run the Streamlit app

   streamlit run app.py

---

Dataset

• Source: spam.csv
• 5572 messages labeled as ham or spam

Author

Sarfras LinkedIn

---

License

This project is open-source and available under the MIT License.