Sentiment Analysis of UMSIDA Student Trust Services Using the Support Vector Machine (SVM) Method

Abstract

The myUMSIDA application supports academic activities at Universitas Muhammadiyah Sidoarjo, but student reviews reveal complaints about its services and facilities. Sentiment analysis is necessary to classify these reviews into positive or negative categories, providing insights to improve service quality.This study uses the Support Vector Machine (SVM) algorithm with a linear kernel, known for its high accuracy, combined with the TF-IDF feature extraction method to enhance text classification. A total of 1,300 reviews from 2023 were processed through labeling, preprocessing, transformation, and classification. Data were split into three scenarios: 70:30, 60:40, and 50:50 for training and testing. Performance was evaluated using accuracy, precision, recall, and F1-Score.The best results were achieved in the 70:30 scenario, with an accuracy of 86.92%, precision of 86.60%, recall of 84.72%, and F1-Score of 85.7%. This study highlights the effectiveness of SVM with a linear kernel and TF-IDF in analyzing sentiment, offering a basis for enhancing the myUMSIDA application's services.