Segmentation and Prediction of Store Performance on the Shopee Marketplace Using a Hybrid Clustering Approach, Spatial Analysis, and Feature Importance

Marketplace platforms have become a central component of digital commerce, particularly in Southeast Asia where Shopee has emerged as one of the dominant e-commerce ecosystems. The increasing number of sellers on the platform intensifies competition and requires data-driven approaches to understand store performance patterns. This study aims to analyze and predict the performance of Shopee stores using a hybrid data mining approach that integrates clustering, spatial analysis, and feature importance evaluation. The dataset consists of 655 Shopee stores collected on February 18, 2026, including attributes such as number of products, chat response rate, follower count, store rating, store tenure, promotional activity, and seller address. K-Means clustering is applied to segment store performance, while spatial analysis examines the geographic distribution of clusters across Indonesian provinces. Furthermore, a Random Forest classifier is used to predict performance categories and identify influential features affecting store competitiveness. The clustering results reveal three distinct store performance groups representing low, medium, and high activity levels. Spatial analysis indicates that provinces with stronger digital ecosystems, particularly West Java and Jakarta, contain a higher concentration of active stores. Feature importance analysis shows that promotional activity, chat responsiveness, and follower count significantly influence store performance classification. The findings contribute to the development of hybrid data mining frameworks for marketplace analysis and provide practical insights for improving seller competitiveness in digital commerce ecosystems.