SMS Spam Classifier

This project implements a machine learning-based spam detection system that classifies SMS messages as either "spam" or "ham" (legitimate) using the Multinomial Naïve Bayes algorithm with TF-IDF feature extraction.

Features

• Text Preprocessing: Comprehensive text cleaning including lowercasing, punctuation removal, tokenization, stopword removal, and lemmatization
• TF-IDF Vectorization: Converts text into numerical features using Term Frequency-Inverse Document Frequency with bigram support
• Multinomial Naïve Bayes: Probabilistic classifier optimized for text classification tasks
• Performance Metrics: Detailed evaluation including accuracy, precision, recall, F1-score, and confusion matrix
• Feature Importance Analysis: Identifies top spam and ham indicator words
• Custom Prediction Function: Easy-to-use interface for classifying new messages

Dataset

SMS Spam Collection Dataset
(https://www.kaggle.com/datasets/uciml/sms-spam-collection-dataset)

• Text messages labeled as spam or ham
• Imbalanced dataset with more legitimate messages than spam
• Preprocessed and cleaned for optimal model performance

Technologies Used

• pandas - Data manipulation and analysis
• numpy - Numerical computing
• matplotlib & seaborn - Data visualization
• nltk - Natural language processing
• scikit-learn - Machine learning algorithms and utilities
  • TfidfVectorizer
  • MultinomialNB
  • train_test_split
  • Performance metrics

Project Structure

.
├── ML_Project_Spam_SMS_Classifier.ipynb  # Main notebook
├── SMSSpamCollection                      # Dataset file
└── README.md                              # Project documentation

Usage

1. Ensure the SMSSpamCollection dataset is in the same directory
2. Open ML_Project_Spam_SMS_Classifier.ipynb in Jupyter Notebook
3. Run all cells sequentially