<![CDATA[In recent years, chatbots have become one of the key innovations in customer service due to their ability to provide fast, accurate, and consistent responses. However, selecting the most suitable machine learning algorithm to accurately classify customer inquiries remains a challenge. This study compares the Naïve Bayes and Random Forest algorithms in intent classification for an Indonesian language-based customer service chatbot. Using a dataset of 26,873 conversations processed through preprocessing stages and TF-IDF vectorization, the evaluation results show that Random Forest achieved an accuracy of 96%, compared to 95% for Naïve Bayes, although both yielded nearly similar precision, recall, and f1-score values. These findings highlight that both algorithms remain relevant, but Random Forest delivers more consistent performance in improving classification accuracy. Practically, this research provides a reference for selecting algorithms in developing customer service chatbots that are more efficient, accurate, and adaptive to user needs, thereby enhancing interaction quality and reducing the workload of human operators.]]>
Copyrights © 2025
