Research Article
Advanced Customer Segmentation Techniques: A Performance Evaluation of Spectral Clustering and Traditional Methods
@INPROCEEDINGS{10.4108/eai.28-4-2025.2358058, author={Venkata Naga Surya Manoj Bulusu and Shanmukha Srinivas Uppada and Appikonda Umesh Krishna Sai and Venkata Krishna Guduru and Karri Sajeev Reddy and Karuna Karri}, title={Advanced Customer Segmentation Techniques: A Performance Evaluation of Spectral Clustering and Traditional Methods}, proceedings={Proceedings of the 4th International Conference on Information Technology, Civil Innovation, Science, and Management, ICITSM 2025, 28-29 April 2025, Tiruchengode, Tamil Nadu, India, Part II}, publisher={EAI}, proceedings_a={ICITSM PART II}, year={2025}, month={10}, keywords={customer segmentation clustering algorithms k-means spectral clustering dbscan gaussian mixture models machine learning evaluation metrics}, doi={10.4108/eai.28-4-2025.2358058} }- Venkata Naga Surya Manoj Bulusu
Shanmukha Srinivas Uppada
Appikonda Umesh Krishna Sai
Venkata Krishna Guduru
Karri Sajeev Reddy
Karuna Karri
Year: 2025
Advanced Customer Segmentation Techniques: A Performance Evaluation of Spectral Clustering and Traditional Methods
ICITSM PART II
EAI
DOI: 10.4108/eai.28-4-2025.2358058
Abstract
Customer segmentation is an essential method in marketing and business analytics that enables companies to customize their strategies to different customer segments. Good segmentation can result in satisfied customers and the best use of resources. The objective of this study is to conduct a comparative study of clustering methods (K-Means, Spectral Clustering, DBSCAN, and GMM) for the segmentation of customers on the Online Retail Dataset from UCI Repository. The imitation dataset is prepared and clustering models are implemented and evaluated by Silhouette Score, ARI, and DBI. The performances highlight that Spectral Clustering is the most consistent algorithm and seems to provide the best results to perform customer segmentation in this application domain, with a Silhouette Score of 0.52 and the best Davies-Bouldin Index (1.20). These results underline the capability of non-linear clustering algorithms in tasks of complex segmentation.
