Sentiment Classification of Aci Application Reviews Using N-Gram Features And Support Vector Machine (SVM) Algorithm

Abstract

The transformation of information technology has created significant opportunities for the application of Natural Language Processing (NLP) in text-based sentiment analysis, particularly in exploring user opinions toward application-based services. This study aims to analyze the sentiment of user reviews of the ACI (Aku Cinta Indonesia) online motorcycle taxi application available on the Google Play Store by applying the N-gram method and the Support Vector Machine (SVM) algorithm. A total of 1,419 reviews were collected, and after data preprocessing and lexicon-based sentiment labeling, 239 final samples were obtained and categorized into positive and negative sentiments. Feature extraction was performed using combinations of unigram, unigram + bigram, and unigram + trigram, with Term Frequency–Inverse Document Frequency (TF-IDF) weighting. Furthermore, the classification process was carried out using a linear kernel Support Vector Machine with an 80:20 split between training and testing data. The experimental results show that the unigram+ bigram model achieved the highest accuracy of 96%, followed by unigram + trigram at 94% and unigram at 90%, with all precision, recall, and F1-score values across the three models exceeding 88%. These findings indicate that the unigram + bigram combination represents word context more effectively than unigram while remaining more efficient than unigram + trigram, thereby improving the sentiment classification accuracy of the SVM model without significantly increasing computational complexity.