<![CDATA[The service quality of a shop can be reviewed through the rating feature which is represented in the form of the number of stars and comments. Customers often provide ratings and comments after purchasing products and receiving service from a shop. However, this is inefficient because potential buyers have to look at the comments one by one. Therefore, Machine Learning is one way to classify these comments. Creating a Service Quality Dictionary is an important point in this classification process because it functions to determine the service class of each comment. The algorithms used include Support Vector Machine, Random Forest, Multinomial Naive Bayes, TF-RF, and Ensemble Learning Bagging. The classification stages include the process of scraping, labeling, data sharing, preprocessing, word weighting, modeling, and model evaluation. Data is classified into two labels, the Service Quality label and the Not Service Quality label. The Service Quality Label will then be reclassified into three labels, namely Positive, Negative and Neutral. Modeling evaluation uses training data with three ratios, namely 90:10, 80:20, and 70:30. From these three ratios, it was found that the 90:10 ratio had better accuracy results, so the modeling evaluation for testing data used the ratio that had the best results from the training data evaluation, namely 90:10 which produced an accuracy of 87,3%.Keywords: Ensemble learning bagging; service quality; text mining; TF-RF.]]>
Copyrights © 2025
