Analisis Sentimen Ulasan Google Maps untuk Layanan Kesehatan Bandar Lampung Menggunakan TF-IDF dan SVM

Abstract

The development of digital technology has encouraged the public to utilize digital platforms like Google Maps as a primary medium for conveying experiences and opinions regarding public services, such as healthcare. These public reviews have become a crucial data source for assessing patient satisfaction and identifying service weaknesses. This study aims to analyze public sentiment towards healthcare services in Bandar Lampung City using machine learning. Support Vector Machine (SVM) was chosen as the text classification method, assisted by Term Frequency-Inverse Document Frequency (TF-IDF) weighting for word feature extraction. The research dataset was obtained through web scraping of 2,278 Google Maps reviews, which were then processed through preprocessing, data labeling, feature extraction, and model evaluation stages. The results of this study indicate a dominance of positive sentiment at 72.5% and negative sentiment at 27.5%. The SVM model with TF-IDF successfully achieved an accuracy of 89% with balanced precision, recall, and F1-score values, categorizing it as superior for text sentiment classification. These findings confirm that aspects of healthcare staff friendliness and service quality were positively rated, while administrative flows, drug distribution, and service speed remain areas of concern. This research utilizes Google Maps reviews as real-time data to evaluate healthcare services in Bandar Lampung City, a previously under-researched area. Sentiment analysis based on SVM and TF-IDF can serve as a data-driven evaluation tool to support strategic decision-making by the Bandar Lampung City Health Office in improving the quality of healthcare services.