ECRO-Kernel DBSCAN: An Optimized Clustering Approach forRobust E-commerce User Segmentation
Informatica · 25.12.2025 · 49 (37)
DOI: 10.31449/inf.v49i37.11124 ↗Kurz gefasst
Fachartikel in der Zeitschrift Informatica, online erschienen am 25.12.2025. Verfasst von Yunyi Zhu und 1 weiteren Autor:innen.
Abstract (Originalsprache)
E-commerce user segmentation is vital for enterprises seeking precise marketing strategies and efficientcustomer management. Traditional clustering algorithms often struggle with noisy data and isolatedpoints, reducing segmentation quality. This study improves e-commerce user segmentation throughadvanced cluster analysis techniques that exclude noisy data and select high-quality initial cluster centers.A novel Efficient Coral Reefs Optimized Kernel Density-Based Spatial Clustering of Applications withNoise (ECRO-Kernel DBSCAN) algorithm is applied for segmentation based on customer behavior. Theresearch implements a data-driven clustering approach using Kernel DBSCAN enhanced with ECROoptimization. Experiments employed a Kaggle e-commerce dataset with 50,000 user records and 12behavioral features, including synthetic noise to mimic real-world variability. Cluster validity wasassessed using the Silhouette, Davies-Bouldin, and Calinski-Harabasz indices. Comparative benchmarksshow the proposed method outperforms traditional K-means and DBSCAN, achieving a 4% improvementin identifying accurate user segments. Analysis revealed five distinct customer groups: Platinum, Gold,Silver, Copper, and Iron, defined by purchasing behavior and engagement levels. Detailed segmentationinforms customized marketing strategies tailored to each group. Effectiveness was further validatedthrough improvements in sales performance and platform user satisfaction. Experimental results indicate94% accuracy, 93% precision, 92% recall, and 94% F1-score, demonstrating superior segmentationrobustness and practical effectiveness compared to conventional methods. This research offers a practicalframework for businesses to optimize customer targeting, enhance engagement, and drive sustainablegrowth through data-driven cluster analysis.
Zitieren
APA
Zhu, Y., & Zhang, X. (2025). ECRO-Kernel DBSCAN: An Optimized Clustering Approach forRobust E-commerce User Segmentation. Informatica, 49(37). https://doi.org/10.31449/inf.v49i37.11124
Vancouver
Zhu Y, Zhang X. ECRO-Kernel DBSCAN: An Optimized Clustering Approach forRobust E-commerce User Segmentation. Informatica. 2025;49(37). doi:10.31449/inf.v49i37.11124
BibTeX
@article{Zhu2025,
author = {Zhu, Yunyi and Zhang, Xuhan},
title = {ECRO-Kernel DBSCAN: An Optimized Clustering Approach forRobust E-commerce User Segmentation},
journal = {Informatica},
year = {2025},
volume = {49},
number = {37},
doi = {10.31449/inf.v49i37.11124}
}
EUVIMED