Aspect Based Sentiment Analysis Review Layanan Telemedicine Menggunakan Algoritma K-Nearest Neighbor

Abstract

Telemedicine is a solution for providing health services by medical personnel to provide long-distance medical services by applying data and communication technology. such as providing diagnostic data, treating, preventing disease and injury, researching and evaluating, and ongoing learning about the importance of health. The variety of health services influences potential users to obtain information and determine which telemedicine services to use. Reviews from users aim to be a solution for companies to classify the speech of telemedicine service users based on predetermined aspects. This research utilizes telemedicine user reviews which are divided into sentiment levels, namely positive, negative and neutral. However, sentiment analysis alone is not enough to carry out classification, further development needs to be carried out so that the system provides more accurate and reliable analysis results, therefore 5000 review datasets from the Halodoc telemedicine service were used. The prepossessing, TFIDF and TF-RF weighting stages were carried out as a comparison in order to find out which word weighting process is more effective to implement using the KNearest Neighbor method. The research was carried out by applying the confusion matrix method so that the final accuracy value was obtained based on 4 aspects. In TF-IDF the highest accuracy used K=11 was 81% and in TF-RF the highest accuracy used K=7 was 78%.