AI-Based Product Clustering For E-Commerce Platforms: Enhancing Navigation And User Personalization

Abstract

With the rapid growth of e-commerce platforms over the years, finding the right products might prove challenging for users. With massive catalogs that are often overwhelming, it might, at times, be harder to navigate, personalize, or even discover the products required by a user on e-commerce platforms. Therefore, in this study, we explore practical AI-based approaches to group similar products together based on their prices, ratings, categories, brands, and user behavior to promote efficacy and combat the aforementioned inefficiencies. First, standard clustering models, K-Means and DBSCAN, are run on e-commerce data to segment products based on the basic features. Since clusters do not reflect how users interact with products, two main upgrades are introduced. First, behavior-like features such as estimated click-through rates and popularity scores are introduced. Second, models such as MiniBatchKMeans and HDBSCAN are implemented for cases where large datasets are to be used, and scalability is a major concern. Since clustering models lack ground truth labels for evaluation, we use the Silhouette Scores and Davies-Bouldin Index to compare the performance of the clustering architectures. With the inclusion of behavioral features in the dataset, it is observed that the performance of the models improves as behavioral features help create tighter, more meaningful groups, and the scalable methods cut down processing time, which makes real-time updates more feasible. This work demonstrates how blending user behavior with flexible clustering techniques can make product organization smarter, ease navigation, and make personalization more effective.