Predicting Future High-Value Customers in E-Commerce: An Explainable Machine Learning Framework for Digital Customer Relationship Management

DOI

10.2991/978-94-6239-830-6_16How to use a DOI?

Keywords

digital economy; e-commerce; customer value prediction; explainable machine learning; customer relationship management

Abstract

Digital customer relationship management increasingly depends on converting transaction histories into actionable forward-looking signals. This study develops a temporally separated and explainable machine learning framework for identifying customers who will enter the top quintile of future spending. Using the public UCI Online Retail dataset, 15 behavioral features are constructed from a 274-day observation window ending on 31 August 2011, and labels are determined from the following 100-day prediction window. Records with missing customer identifiers, cancellations or returns, and non-positive prices are removed, while customers with only one valid historical purchase are retained. Logistic regression, random forest, and XGBoost are evaluated by stratified five-fold cross-validation and by rolling-origin tests. The random forest achieves mean ROC-AUC of 0.861 and PR-AUC of 0.692. Its ROC-AUC remains 0.896, 0.861, and 0.846 under top-10%, top-20%, and top-25% definitions, respectively, and reaches 0.861 and 0.852 in two out-of-time tests. SHAP analysis identifies cumulative monetary value, product diversity, recent revenue, purchase frequency, and average basket value as the main predictive signals. The results support a transparent customer-prioritization workflow whose conclusions are not dependent on a single value threshold or a random cross-validation split.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Yaxin Chi
PY  - 2026
DA  - 2026/11/14
TI  - Predicting Future High-Value Customers in E-Commerce: An Explainable Machine Learning Framework for Digital Customer Relationship Management
BT  - Proceedings of the 6th International Conference on Internet Finance and Digital Economy (ICIFDE 2026)
PB  - Atlantis Press
SP  - 192
EP  - 202
SN  - 2352-5428
UR  - https://doi.org/10.2991/978-94-6239-830-6_16
DO  - 10.2991/978-94-6239-830-6_16
ID  - Chi2026
ER  -