This document describes a study that aimed to develop a real-time traffic classification system using Twitter data in Yogyakarta, Indonesia. The study collected over 110,000 tweets, preprocessed them, extracted features, and used machine learning classifiers like Naive Bayes, Support Vector Machine, and Decision Tree to classify tweets as related to traffic or not. Experimental results showed that for balanced datasets, SVM achieved the best performance of 99.77% accuracy, while for imbalanced datasets, SVM also performed best with 99.87% accuracy. The study demonstrates the potential of using social media data for real-time traffic anomaly detection.