GOJEK DATA ANALYSIS THROUGH TEXT MINING USING SUPPORT VECTOR MACHINE (SVM) AND K-NEAREST NEIGHBOR (KNN)

Abstract

The main focus of this research is to apply and test the effectiveness of SVM and KNN methods in Gojek data text analysis. This research will examine how the two methods can classify user comments and feedback and identify data sentiment analysis at the same time practically help Gojek understand user needs and improve service quality. The data obtained through scrapping is categorized into positive and negative sentiment. Data is taken from Gojek application user reviews throughout the year 2022 with a total of 1148 sentiment data with a percentage of 80% training data and 20% testing data. Evaluation of model performance using Confusion Matrix and AUC-ROC Curve shows that SVM is more effective than KNN, with accuracy on training data of 92.55% for SVM and 81.71% for KNN, as well as accuracy on testing data of 82.40% for SVM and 77,09% for KNN.