Customer Lifetime Value Modelling with Gradient Boosting

Main Article Content

Arjun Sirangi

Abstract

Customer Lifetime Value (CLV) is a critical metric for businesses to forecast long-term customer profitability and optimize resource allocation. Traditional CLV models, reliant on heuristic or probabilistic approaches, often fail to capture complex customer behavior patterns in highdimensional datasets. This paper proposes a gradient boosting framework for CLV prediction, integrating survival analysis, advanced feature engineering, and hyperparameter optimization. Empirical results demonstrate a 22% improvement in prediction accuracy over conventional methods, validated through quantile scoring and customer decile analysis. Challenges such as data sparsity, ethical bias, and non-stationary environments are addressed, alongside future directions in hybrid deep learning and causal inference.

Article Details

Section
Articles