Spam Text Classifier

This project implements a text classification model trained on SMS/text message data to distinguish between spam and non-spam (ham) messages. The model applies natural language preprocessing and TF-IDF vectorization before training a Support Vector Classifier to make binary predictions on unlabeled message data, outputting results to a predictions.csv file.

Capabilities

• Text Preprocessing: Converts text to lowercase, removes punctuation, and strips digits to normalize input data
• TF-IDF Vectorization: Converts cleaned text into numerical feature vectors using TfidfVectorizer with English stop-word removal and a 5,000-feature limit
• Support Vector Classification: Trains a LinearSVC model to classify messages as spam or ham based on learned text patterns
• Prediction Export: Generates a predictions.csv file with binary labels (TRUE for spam, FALSE for ham) for each test message

Usage

Terminal MAC Run Script

bash run.sh

or directly:

python3 model.py

Required input files:

• data_train_hw4_problem1.csv — Labeled training data with spam and text columns
• data_test_hw4_problem1.csv — Unlabeled test data with text column

Output:

• predictions.csv — Binary classification results with spam column (TRUE or FALSE)

Use Cases

Personal Message Filtering

Evaluate SMS or text message datasets to identify unwanted spam messages for personal or organizational message triage.

Research on Text Classification

Serve as a baseline or reference implementation for studying linear SVM-based text classification pipelines using TF-IDF features.

Research Purposes

Designed for research purposes. The repository includes a dummy dataset for demonstration and testing purposes. Penn State University (PSU), IST 557 Data Mining. Fall 2025.