<![CDATA[Sentiment analysis aims to identify user opinions about products on marketplaces such as Shopee and Tokopedia. This study classifies product review sentiment using Naive Bayes (NB) and Support Vector Machine (SVM). The dataset underwent text preprocessing, including case folding, tokenization, stopword removal, and stemming, then was represented using TF-IDF. The results show that Support Vector Machine (SVM) achieved the highest accuracy of 94.54%, but had a very low negative class recall (5.71%), indicating a strong bias towards the majority class. In contrast, Naïve Bayes (NB) recorded a lower accuracy of 67.88%, but showed more balanced performance with a negative class recall of 48.57%. Conversely, NB provided more balanced performance between positive and negative classes despite its slightly lower accuracy. These findings emphasize the importance of considering class imbalance in sentiment analysis, especially for applications that require consumer complaint detection. This research is expected to serve as a reference for the development of automatic sentiment analysis systems on marketplace platforms with a focus on performance balance between classes.]]>
Copyrights © 2026
