Perbandingan Kinerja Support Vector Machine dan Neural Network dalam Klasifikasi Sentimen Ulasan Pengguna Aplikasi AppSheet

Abstract

User reviews on digital platforms such as Google Play Store provide valuable information regarding user perceptions and experiences toward mobile applications. However, the large volume of reviews makes manual analysis inefficient. Sentiment analysis is widely used to automatically extract opinions from textual data. This study aims to compare the performance of Support Vector Machine (SVM) and Neural Network algorithms in classifying sentiment from AppSheet user reviews. The research employed a quantitative experimental approach using 1,000 user reviews collected through web scraping. The data were processed through several preprocessing stages including cleaning, case folding, tokenization, stopword removal, normalization, and stemming. Sentiment labeling was conducted using a lexicon-based approach with the Indonesian Sentiment Lexicon (InSet). Feature extraction was performed using TF-IDF Vectorizer, producing 1,277 features. The dataset was divided into training data (80%) and testing data (20%). Model performance was evaluated using accuracy, precision, recall, and F1-score. The results indicate that the SVM model achieved an accuracy of 86%, while the Neural Network model achieved 89%. These findings demonstrate that Neural Network performs better in capturing non-linear sentiment patterns in Indonesian-language review data. The study concludes that the integration of lexicon-based labeling and machine learning methods can effectively improve sentiment classification performance.