The Indonesian e-commerce market has experienced extraordinary growth, driven by increasing internet penetration and smartphone adoption, which necessitates advanced data analysis for competitive advantage. Clustering is a crucial data mining technique used to group products based on similar characteristics, providing in-depth insights into product performance. Previous studies often focused on single performance metrics, overlooking the nuances of combining multiple variables. This study aims to address this gap by implementing and comparing the K-Means and K-Medoids clustering algorithms on Tokopedia product data using a combination of numerical attributes: Price, Customer Rating, Number Sold, and Total Review. The methodology involved data preprocessing, Min-Max Scaling for normalization, and using the Elbow Method to determine the optimal number of clusters, which was found to be K=2. The clustering quality was rigorously evaluated using the Davies-Bouldin Index (DBI) and Silhouette Score. The results demonstrate that K-Means exhibits superior performance, achieving a lower DBI of 0.5717 and a higher Silhouette Score of 0.6012, compared to K-Medoids (DBI: 0.5870; Silhouette Score: 0.5857). Furthermore, K-Means proved significantly more efficient computationally, with an execution time of 0.0947 seconds versus 0.1622 seconds for K-Medoids. The main conclusion is that K-Means is more effective in creating compact and clearly separated clusters. This research contributes a valuable analytical framework for e-commerce managers to comprehensively understand product profiles, guiding more effective marketing and recommendation strategies.
Malik et al. (Fri,) studied this question.