Optimalisasi Sistem Chatbot Layanan Pelanggan E-Commerce melalui Kombinasi Rule-Based dan Algoritma K-Nearest Neighbor (KNN)

Abstract

This research aims to design and evaluate a hybrid chatbot prototype that combines a rule-based approach with the K-Nearest Neighbor (KNN) algorithm to classify customer satisfaction levels. This system was developed as an automated customer service solution for the e-commerce sector, which demands speed, efficiency, and personalization. By using numerical features such as information comprehension (X1), chatbot rating (X2), and conversation duration (X3), the KNN algorithm determines whether a user is satisfied or not after an interaction session. The classification results indicate that some "Dissatisfied" predictions occurred in cases with high information comprehension but also long conversation durations. This suggests that customers can still be dissatisfied if the service time is perceived as too long, even if the content of the answer is clear. Although there is a margin of error in the user experience aspect, the system demonstrates a very significant impact on operational efficiency. The implementation of the hybrid chatbot was able to reduce customer service costs by up to 50%, with an estimated payback period of 4.8 months. The system provides significant added value for businesses, particularly in terms of cost savings and scalability, although there is still room for improvement in the quality of interaction.