Sentiment Analysis on Jobseekers Application in Google Play Store (KitaLulus)

Abstract

Sentiment analysis on user reviews of the KitaLulus application in Google Play Store aims to assess user feedback and classify sentiment into positive and negative categories. This study applies and compares the performance of Support Vector Machine (SVM) and Naïve Bayes classifiers in sentiment classification. Data was collected using the Google-Play-Scraper API, yielding 16,488 reviews, which underwent preprocessing, including case folding, tokenization, stopword removal, and lemmatization. The dataset was divided into training (80%), validation (5-fold cross-validation), and testing (20%) sets. During validation, the training data was further split, using 64% for training and 16% for validation in each iteration. The results indicate that SVM outperforms Naïve Bayes, achieving 93.99% accuracy, 97% precision, 92% recall, and an F1-score of 94%, while Naïve Bayes achieves 89.89% accuracy, 94% precision, 87% recall, and an F1-score of 90%. These findings demonstrate that SVM provides a more balanced classification performance, making it a more suitable approach for sentiment analysis in this context. This research contributes to a better understanding of user sentiment and provides valuable insights for improving the KitaLulus application.