Product and Store Recommendation System Using K-Means Clustering and Hybrid Filtering on Marketplace

Abstract

The development of information and communication technology has driven significant changes in the digital business landscape, particularly in the e-commerce sector. Marketplaces have become crucial platforms for connecting consumers with product providers, including supporting the growth of Micro, Small, and Medium Enterprises (MSMEs). As transaction volumes and product diversity continue to increase, new challenges have emerged in providing consumers with relevant product recommendations. This study aims to develop a product and store recommendation system by combining K-Means Clustering for customer segmentation and Hybrid Filtering to enhance recommendation accuracy. The system was developed using an experimental approach based on software engineering, with historical transaction data from the CV. Talongka Jaya marketplace as the primary data source. Customer segmentation resulted in five clusters based on purchasing behavior patterns, such as transaction frequency and product category preferences. These clustering results were then used to tailor product and store recommendations to the characteristics of each segment. The recommendation system was built by integrating Collaborative Filtering and Content-Based Filtering with optimal weights of 0.7 and 0.3, respectively. Evaluation using 5-fold cross-validation demonstrated that Hybrid Filtering achieved a Precision of 0.78 and an F1-Score of 0.74, outperforming single-method approaches. These findings confirm that the integration of clustering and hybrid filtering is effective in enhancing service personalization and improving users’ shopping experience. This research makes a significant contribution to the development of data mining-based recommendation systems for MSME marketplaces, although there remains room for further improvement through the integration of real-time data and deep learning-based sequential recommendation methods.