Sentiment Analysis of Application User Feedback in Bahasa Indonesia Using Multinomial Naive Bayes

User feedback can be used as a tool for application developers to find out and understand users' needs, preferences, and complaints. Developers need to identify problems that arise from the user-given feedbacks, which is very difficult to do, considering the amount of feedback received every day. Reading and classifying every feedback takes a long time, and it is very ineffective. In order to overcome this problem, a sentiment analysis system based on the Multinomial Naïve Bayes classification algorithm was built to determine whether a user-feedback has a positive or negative sentiment. Naïve Bayes algorithm is generally used for classification because it is straightforward and effective. Based on previous research, the Multinomial Naïve Bayes algorithm gives off the best performance compared to other traditional machine learning algorithms. This study aims to implement the Multinomial Naïve Bayes classification algorithm on a web application and calculate the accuracy of class predictions made by the system. Based on the results of several test, the accuracy of class predictions which were evaluated using confusion matrix, shows that the model with training-testing comparison of 70:30, balanced datasets, and oversampling each dataset by 30% produces the best performance, with 71.6% for accuracy, 76.92% for precision, 61.73% for recall and 68.49% for F1 score.