Web growth contributes greatly to user-generated content such as user feedback, opinions and reviews. The construction of the Jakarta-Bandung High Speed Train is both an icon and a momentum for Indonesia to modernize mass transportation in an era of continuous progress. Sentiment analysis is one of the text-based research field solutions suitable for addressing satisfaction issues based on user reviews. In this research, the system will be made with review sentences from users and produce output in the form of positive and negative classes. The method used by the author is classification using the Support Vector Machine (SVM) method and Word2Vec extraction features. In addition, a comparison of the accuracy value between the Support Vector Machine method, Naïve Bayes method and TF-IDF extraction features is carried out. The data studied came from several news websites containing user reviews of the Jakarta-Bandung High Speed Train. This method is used because it represents words in a vector, besides that the training process is faster when compared to other extraction features. This research resulted in the performance of accurasy, precision, recall, and f1-score, namely accurasy of 82.74%, precision of 75.68%, recall of 97.67%, and f1-score of 85.28%. These results were obtained using the best tuning hyperparameters, namely ('C': 10, 'gamma': 0.1, 'kernel': 'rbf'). Then in the second scenario a comparison is made with the Naïve Bayes method. It was found that the accuracy of the Support Vector Machine method using the TF-IDF extraction feature obtained better and stable performance results than the other three performance results, which amounted to 86.90%. So the author concludes that the Support Vector Machine method using the TF-IDF extraction feature is better when compared to the Naïve Bayes method and the Word2vec extraction feature.