Sentiment analysis of mass rapid transit jakarta using naïve bayes classifier and rule-based opinion target detection on Twitter

Abstract

Jakarta Mass Rapid Transit (MRT) is a national project in Indonesia that has been operated since the beginning of 2019 for phase 1 and still under further development. People's opinions posted on Twitter concerning Jakarta MRT could become an evaluation material, viewed from the sentiment score and object of the opinion. To analyze the sentiment score, opinion sentiment classified using the naive Bayes classifier, followed by rule-based target opinion detection. The classification process is using bag of words (BoW) features and lexicon-based features. In the weighting process of the Lexicon-based features and detection of the target opinion, POS tagging is employed to get the word class. In determining the target opinion object, the POS tagging result is used to do chunking that has specific rules, specific to noun-phrase (NP) tags. Therefore, the obtained sentiment class and object become the target in the opinion. Using Naïve Bayes with the bag of words features and lexicon-based, we achieve precision 0,92, recall 1,0, f-measure 0,95, and accuracy 0,92. The results of the rule-based target opinion detection are 0.78, 0.85, 0.79, and 0.75 for precision, recall, f-measure, and accuracy, respectively.