COMPARING ALGORITHM FOR SENTIMENT ANALYSIS IN HEALTHCARE AND SOCIAL SECURITY AGENCY (BPJS KESEHATAN)

Abstract

Twitter is a social media that can be used to express opinions and exchange information quickly with individuals and institutions such as the Healthcare and Social Security Agency (BPJS Kesehatan). Every word that a Twitter user utters has meaning and stellar emotion. This meaning can be reached through the process of sentiment analysis. Sentiment analysis is the process of understanding and classifying emotions such as positive or negative or complaining or not complaining. This study classifies tweet data related to BPJS Health services into two classifications, namely complain and no complain. Using 1,000 data from Twitter written on the BPJS Kesehatan Twitter account. In text mining, to build a classification, the transform case, tokenize, token filter by length, stemming and stopword techniques are used. Gataframework is used to assist the preprocessing and cleansing process. Rapidminer was used to create sentiment analysis in comparing three different classification methods of the Twitter data. The method used is the Nave Bayes algorithm and the Naïve Bayes algorithm with the addition of a Synthetic Minority Over-sampling Technique (SMOTE) feature and the Naïve Bayes algorithm with an SMOTE feature that is optimized with Adaboost. The Naïve Bayes algorithm is added with the SMOTE feature which is optimized with Adaboost to get the best value with an accuracy value of 69.11%, precision 69.93%, recall 68.89% and AUC 0.770