SENTIMENT ANALYSIS ON TWITTER ACCOUNT USING NAIVE BAYES CLASSIFIER ALGORITHM Case Study: Indonesia Healthcare and Social Security Agency (BPJS Kesehatan)

Abstract

BPJS Kesehatan is organizing the health care insurance for all Indonesian people. By 2018, the number of participants of BPJS Kesehatan reached 196,662,064 people. A large number of these users make BPJS Kesehatan must provide services in the form of feedback. One uses media is Twitter. Information obtained from any tweets, can be used as a tool of policy makers and this can be done by using sentiment analysis. At sentiment analysis, a classification method that can be used is Naive Bayes classifier algorithm. Naive Bayes classifier algorithm is a classification method that is rooted in the Bayes theorem. In this paper we show a system of sentiment analysis BPJS twitter account with a Naive Bayes classifier algorithm. Naive Bayes classifier algorithm consists of two stages. The first stage is to set the sample document training (training data) and the second stage is the process of classifying documents of unknown category (class). The system uses a method Naive Bayes classifier algorithm for classification. Phase to be conducted before the classification is preprocessing. Stages in the preprocessing consists of a folding case, normalization features, the emoticons convert, convert negation, tokenizing, stemming and stopword removal. Tweets that have passed the stage of preprocessing will be classified into positive opinion or negative opinion and displayed in a pie chart. Based on testing, the results tweets classification accuracy is 88% with precision positive 85%, precision negative 75% and precision neutral 92%.