Abstract

Abstract The increasing shift towards online shopping has enabled businesses to collect and analyze customer clickstream data to understand consumer behavior. Traditional machine learning algorithms struggle to segment high and low-profit customers effectively. This study proposes an advanced clustering-based approach incorporating Partitioning Around Medoids (PAM), Gower Distance Matrix, and Kruskal-Wallis analysis to segment online shoppers based on revenue generation. A dataset similar to the UK ASOS dataset was utilized from the Kaggle repository for analysis. The findings suggest that this approach effectively segments consumers, allowing businesses to optimize marketing strategies for high-revenue customers.The primary goal of this research is to enhance the accuracy of consumer segmentation using clickstream data, enabling e-commerce platforms to identify and target highvalue customers efficiently. By leveraging the PAM clustering technique, the study ensures better interpretability of clusters, while the Gower Distance Matrix effectively processes mixed data types. The Kruskal-Wallis statistical test is applied to determine revenue distribution across clusters, offering valuable insights into customer purchasing behaviors. Experimental results demonstrate that clusters with higher order frequencies contribute significantly to revenue, validating the effectiveness of the proposed methodology. The graphical visualization of customer segments further aids in understanding the distribution and characteristics of different consumer groups. This research contributes to the field of data-driven consumer analytics by offering a scalable and efficient approach to online shopper segmentation, paving the way for more personalized and strategic marketing initiatives.