Sentiment Analysis and Automatic Response Generation for E-Commerce Comments
Ayşe Macit, Seda Postalcıoğlu · Natural and Applied Sciences Journal · 2025
This study addresses the use of machine learning techniques for automatic classification of product reviews on e-commerce sites and generating appropriate responses. It was carried out with approximately 15,000 data labeled as positive, negative and neutral obtained from the "E-Commerce Product Reviews" data set. The TF-IDF vectorization method, which is a text mining technique, was used in the study. Multinomial Naive Bayes, Support Vector Machine, Random Forest, Logistic Regression techniques were used for sentiment analysis. As a result of the studies, the accuracy values of Multinomial Naive Bayes, Support Vector Machine, Random Forest, Logistic Regression algorithms show successful results as 87%, 88%, 85% and 88%, respectively. As a result, it was concluded that automatic comment analysis tools can be effective in improving customer relations for e-commerce sellers.