Sentiment analysis of student responses related to information system services using Multinomial Naïve Bayes (Case study: Telkom University)

Abstract

Current technological developments affect the types of information services provided, especially information system services that always related to the internet network. To improve existing services, user responses are needed regarding services provided, but if the response is very large, it will be difficult to know whether the services provided are good or not. By taking the right approach, the sentiment's response can be analysed quickly and automatically. Sentiment analysis is done by classifying responses into positive and negative classes, the classification method is Multinomial Naïve Bayes (MNB). Before the data is known for the sentiments, the data is manually labeled, then the data goes through the preprocessing stage, cross validation, feature extraction and then sentiment classification with the MNB classifier. Classification process using MNB method on positive and negative responses, obtaining the highest negative response results in iGracias which is a detailed response of the information system sub-service by 44.27%. The MNB method using lemma with the conditions for preprocessing has a good average accuracy of 83.24% for information system sub-services and 79.24% for internet sub-services.