Sentiment analysis is the process of understanding, extracting and processing data textual automatically to get the sentiment information contained in a sentence of opinion expressed in the form and group them into two groups: positive opinions and negative opinions. The data is used from previous project with lexicon, based on positive and negative sentiments. Several processes in this project are divided data into training and test data with ratio 90:10, calculated the word weight with Term Frequency-Inverse Document Frequency (TF-IDF), use Support Vector Machine on training data to produce the classification model then tested on test data.