Application of Sentiment Analysis on Product Reviews of the Binjai Langkat Buket Shop to Improve Customer Service using the Naive Bayes Method

Abstract

In the digital era, customer reviews are spread across various social media and e-commerce platforms, posing a challenge for micro-businesses to evaluate sentiment efficiently. This study aims to develop an automated sentiment analysis system for product reviews of Toko Buket Binjai Langkat using the Term Frequency-Inverse Document Frequency (TF-IDF) method for feature extraction and the Naive Bayes algorithm for positive, negative, and neutral sentiment classification. Data were collected through web scraping techniques and processed with preprocessing stages such as case folding, stopword removal, and stemming. The model was trained and tested with a 70:30 data split and evaluated using accuracy, precision, recall, and F1-score metrics. The test results showed that the model's accuracy reached 91%, with the best performance on positive sentiment (precision 0.89, recall 1.00, F1-score 0.94), but there were limitations in detecting negative sentiment (recall 0.42) due to data imbalance. This study provides practical contributions for micro-businesses in understanding customer opinions and formulating data-driven service improvement strategies.