Comparison of Kernel Support Vector Machines in Conducting Sentiment Analysis Review of Buying Chips on the Shopee E- Marketplace in Indonesian

Abstract

Small and Medium Enterprises (SMEs) have an important and strategic role because they have advantages in fields that utilize natural resources and are labor intensive. Therefore, E-Marketplace's utilization positively influences improving marketing and sales performance in SMEs. Chips are one of the popular SMEs processed products and vary from cassava chips, taro chips, banana chips, spinach chips, etc. In addition to stores, we can find this product on the Shopee e-marketplace. To improve and maintain the quality of chips as an SME product, it is necessary to analyze sentiment. This study conducted a sentiment analysis of the review of buying chips on the Shopee E-marketplace using the NLP approach, which was then carried out by classification modeling using SVM. This research tries to find the best model of the SVM kernel among Linear, Polynomial, Gaussian RBF, and Sigmoid kernels. Based on the experiments conducted, the Linear kernels provide the best performance. The result of linear kernel accuracy and recall is 89.60%. While the linear precision kernel value is 88.8%, then the Fl-Score value is 88.60%. In the end, the linear kernels in the SVM Algorithm have good potential to be used in the development of various systems related to detecting user reviews on the Indonesian-language Shopee E-Marketplace.