This document discusses sentiment analysis on Twitter data using machine learning techniques. It begins with introducing sentiment analysis and its goals for Twitter data, including determining if tweets convey positive, negative, or neutral sentiment. It then outlines the challenges of analyzing Twitter data and its approach, which includes downloading tweets, preprocessing, feature extraction, and using an SVM classifier. It finds its feature-based model performs better than the baseline model, with an accuracy of 57.85% and F1 score of 61.17% for sentence-level sentiment classification. The tools used include Python, Java, LIBSVM, NLTK, and the Twitter API.