Sentiment Analysis on Satusehat Application Using Support Vector Machine Method

Abstract

Sentiment analysis is important in language processing and machine learning. SVM is proven to classify positive and negative sentiments with high accuracy effectively. SatuSehat application provides users with various health services and medical information, previously known as the PeduliLindungi Application. Once, this application was used to handle vaccination history used in the new normal era. Along the way, many problems arose due to the immaturity of the application after it was launched, which resulted in many user reviews being given through the Google Play Store application. Therefore, this study aims to determine SVM's performance in classifying user reviews of the SatuSehat application into positive and negative sentiments and to show visualization to find out the most frequent words from user reviews. Based on the research results, 25,000 data were divided into 18,359 negative class data and 6,641 positive class data. At the SVM classification stage, it produces a negative sentiment of 73.4% and a positive sentiment of 26.6%. In addition, the results of the SVM accuracy test obtained a result of 91% with a positive sentiment, namely having a precision test of 92%, a recall of 71%, and an f1-score of 80%, while for negative sentiment, namely having a precision test of 90%, a recall of 98% and f1-score of 94%. The visualization results found that the topics often appearing in positive reviews are good and sometimes great. In contrast, the negative reviews are update, difficult, strange, login, and bug.