Research Article
Customer Lifetime Value Prediction Model Focus on High-Value Customer Reliability Analysis
@INPROCEEDINGS{10.4108/eai.22-5-2026.2365222, author={Bowen Xu}, title={ Customer Lifetime Value Prediction Model Focus on High-Value Customer Reliability Analysis}, proceedings={Proceedings of the 4th International Conference on Image, Algorithms, and Artificial Intelligence, ICIAAI 2026, 22-24 May 2026, Singapore, Singapore}, publisher={EAI}, proceedings_a={ICIAAI}, year={2026}, month={8}, keywords={Customer Lifetime Value machine learning logarithmic transformation customer analytics}, doi={10.4108/eai.22-5-2026.2365222} }- Bowen Xu
Year: 2026
Customer Lifetime Value Prediction Model Focus on High-Value Customer Reliability Analysis
ICIAAI
EAI
DOI: 10.4108/eai.22-5-2026.2365222
Abstract
Customer Lifetime Value (CLTV) prediction is a fundamental task in customer analytics. It is a challenging problem due to its highly skewed, long-tailed and unbalanced distribution, especially for high-value customers. This research uses three machine learning models: Random Forest, eXtreme Gradient Boosting (XGBoost) and Gradient Boosting to generate the future six-month CLTV and test the stability of these models in the high-value customer dataset. During feature engineering, a logarithmic transformation and inverse restoration strategy are used to reduce distributional skewness and improve model performance. Eventually, Gradient Boosting delivers the strongest overall performance, achieving R2 = 0.9613, RMSE = 459.7925 and MAE = 101.8458 on the full dataset. Gradient Boosting still maintains strong performance on high-value segments with R2 = 0.9406 on top25% dataset and R2 = 0.9161 on top10% dataset.
