Comparison Of Support Vector Machine Radial Base And Linear Kernel Functions For Mobile Banking Customer Satisfaction Analysis

Abstract— Banking services using mobile banking
applications, including Indonesian state bank (called BRI). A
study on feedback regarding BRI services based on mobile
applications was done. In order to compete with other banks, that
is used to enhance and modernize the quality of BRI services
provided to clients. Based on phenomena that occur in these
situations. This study aims to classify comments from users of the
BRI Mobile Banking Application on Google Play services into
positive and negative comment sentiments. In this study, the
Support Vector Machine (SVM) technique is utilized to determine
between positive or negative reviews. The sentiment analysis of
BRI google play data was carried out by comparing the Radial
Basis Function (RBF) kernel function and the Linear kernel. As
well as the experiment of adding feature selection, parameters, and
n-grams for a period of two years, from January 1st,, 2017 to
December 31st, 2018. The results of the study using the k-fold
cross-validation test, the precision value of the SVM kernel linear
is 90.80 percent and the SVM kernel RBF is 90.15 percent. In the
RBF kernel, there are 1,816 positive classes and 1,455 negative
classes. While the Linear kernel obtained a positive class of 1,734
and a negative class of 1,637.
Keywords— Sentiment Analysis, Support Vector Machine,
Kernel RBF and Kernel Linear