SENTIMENT DETECTION OF SHOPEE E-COMMERCE APPLICATION REVIEWS USING NATURAL LANGUAGE PROCESSING AND SUPPORT VECTOR MACHINE

Abstract

Shopee, as a leading e-commerce platform in Indonesia, receives millions of reviews that reflect both user satisfaction and complaints. These reviews serve as a crucial source of strategic information for improving service quality; however, their vast quantity necessitates accurate and automated analysis. This study aims to develop a sentiment detection system for Indonesian-language Shopee reviews by utilizing Natural Language Processing (NLP) techniques and the Support Vector Machine (SVM) algorithm. The methodology includes text preprocessing stages (data cleaning, case folding, tokenization, stopword removal, and stemming using Sastrawi), feature extraction using Term Frequency-Inverse Document Frequency (TF-IDF) and Bag of Words (BoW) approaches. The SVM model is then trained to classify reviews into positive or negative sentiments. The results demonstrate that the model using TF-IDF features outperforms the BoW approach, achieving an accuracy of 93.3%, with precision and recall of 93%, and a high F1-score. These findings reinforce the effectiveness of combining NLP and SVM for analyzing Indonesian-language texts and highlight the critical role of preprocessing stages in enhancing model performance. In conclusion, the developed system offers a practical solution for automatically monitoring user perceptions, supporting data-driven decision-making, and strengthening the competitiveness of e-commerce platforms in an increasingly competitive digital era.