Sentiment Analysis of Shopee App Usage Reviews Using Metaheuristic Algorithm Optimization

Sentiment analysis is widely used to classify opinions expressed in textual data; however, many previous studies employing Support Vector Machine (SVM) do not explicitly address parameter optimization, which can lead to suboptimal classification performance. To address this research gap, this study integrates Particle Swarm Optimization (PSO) to optimize key SVM parameters for sentiment analysis of Shopee application reviews. The dataset consists of 1,000 Indonesian language user reviews collected from the Google Play Store between January and August 2025. Text preprocessing was conducted prior to feature extraction using the Term Frequency–Inverse Document Frequency (TF-IDF) method. PSO was applied to optimize the kernel parameters, penalty parameter (C), and gamma value of the SVM model. The results demonstrate that PSO-based optimization significantly improves classification performance, increasing accuracy from 86.0% to 90.0%, precision from 86.9% to 90.6%, recall from 87.6% to 91.2%, and F1-score from 87.2% to 90.9%. Additionally, positive reviews are dominated by keywords such as “cheap,” “fast,” and “free shipping,” while negative reviews frequently contain terms such as “error” and “slow.” These findings confirm that PSO effectively enhances SVM performance and provides reliable insights into user sentiment toward e-commerce applications.