Abstract
Purpose: This paper proposes a data-driven framework for segmenting retail customers using machine learning, identifying complementary and substitutive product relationships within segments via cross-elasticity of demand, and applying main path analysis to reveal key bundling and substitution patterns. Design/methodology/approach: Machine learning is used for customer segmentation; cross-elasticity of demand (XED) identifies product bundling and substitution patterns within segments; a weighted product network is constructed; and main path analysis (MPA) determines key product chains. Findings: Applied to a large retail transaction dataset, the framework calculates XEDs, defines arc weights by relationship likelihood and constructs a product network; MPA then identifies the key product bundles and substitution patterns most frequently occurring in customer purchases. Practical implications: The framework equips retailers to optimize product offerings and marketing strategies by identifying key product relationships, thereby enhancing targeted promotions, improving inventory management and boosting sales and customer satisfaction. Originality/value: The paper integrates machine learning, XED and MPA to model product relationships as a network problem, demonstrating how real-world transaction data yields new insights into bundling and substitution patterns that support retail marketing, pricing and inventory decisions.

This publication has 51 references indexed in Scilit: