Optimizing Aspect-Based Sentiment Analysis for Kyai Langgeng Park Using PSO and SVM

Abstract

This study aims to analyze aspect-based sentiment on Taman Kyai Langgeng tourism reviews, focusing on three main aspects: price, service, and facilities. This study combines Particle Swarm Optimization (PSO) method for feature selection and Synthetic Minority Over-sampling Technique (SMOTE) to handle data imbalance, which is a novel approach in aspect-based sentiment analysis. A total of 827 review data were retrieved from the Google Maps platform and manually labeled. This method resulted in significantly improved sentiment classification accuracy over the model without optimization. After the application of PSO and SMOTE, the model accuracy for the price aspect increased from 91.56% to 94.28%, the service aspect from 89.75% to 92.85%, and the facility aspect from 79.51% to 88.88%. The results of this study show that the combined PSO and SMOTE approach not only improves the accuracy, but also the consistency of sentiment classification on various aspects. These findings provide deep insights for tourism managers in identifying strengths and weaknesses based on visitor reviews.

Keywords

• aspect-based sentiment analysis
• pso
• smote
• svm