ANALYSIS OF GOOGLE MAPS REVIEW SENTIMENT ON BLIMBING MARKET FACILITIES WITH SVM

Traditional markets play a crucial role in the local economy, especially in developing countries like Indonesia. Although markets and traders have officially developed, the quality of facilities in traditional markets remains a challenge, with various issues affecting visitor comfort and experience. This study proposes a sentiment analysis of public opinions on the facilities of Blimbing Traditional Market using the Support Vector Machine (SVM) method. The aim is to understand the implementation and results of sentiment classification using SVM. The data used is sourced from user reviews on Google Reviews. The analysis process includes data collection, automatic sentiment labeling using TextBlob, and data preprocessing, including case folding, text cleaning, tokenization, stemming, and stopword removal. SVM with various kernels (linear, polynomial, RBF, and sigmoid) was evaluated with train-test data split ratios ranging from 90/10 to 10/90. The results show that the linear kernel achieved an accuracy of 98.07% with an 80/20 ratio, poly 96.15% with a 90/10 ratio, RBF 97.69% with a 90/10 ratio, and sigmoid 98.46% with a 90/10 ratio. Suggestions for future research include exploring more advanced preprocessing techniques and other algorithms such as Random Forest, Naive Bayes, or Neural Networks to improve sentiment analysis performance.