A machine learning approach to decision support for e-commerce adoption initiatives by small and medium enterprises

Purpose This study aims to support small and medium enterprises (SMEs) in making informed e-commerce adoption decisions by identifying key business constructs and developing a machine learning–based decision support tool. Design/methodology/approach A framework grounded in the technology–organization–environment (TOE) framework and perceived strategic value (PSV) principle is proposed. An extended multi-objective micro genetic algorithm (MmGA) is employed to identify influential business constructs, including perceived benefits, perceived obstacles, competitive pressure, government support, operational support, firm size and business models. Classification models are trained using survey data collected from manufacturing SMEs. Findings The proposed MmGA-based model provides e-commerce adoption predictions with useful insights into technological readiness, organizational capability and managerial intent affecting adoption decisions. Research limitations/implications The study relies on self-reported survey data, which may introduce recall and social desirability bias and limits generalizability across regions and time. In addition, the scope focuses on e-commerce pre-adoption stages. Future work can extend the proposed multi-objective model to post-adoption factors such as scalability and customization. Practical implications The developed decision support tool assists SME managers and policymakers in evaluating adoption readiness and prioritizing strategic factors before entering e-commerce activities. Originality/value While the TOE and PSV frameworks are well-established individually, the primary novelty of this study lies in their integration into a multi-objective machine learning paradigm. Unlike traditional regression models that analyze constructs in isolation, our approach treats e-commerce adoption as a multi-objective optimization problem (MOP). This allows for the simultaneous optimization of predictive performance and model parsimony, aiming to undertake the non-linear trade-offs between objective organizational readiness (TOE) and subjective strategic intent of decision-makers (PSV). The extended MmGA model serves as a useful prediction tool with a parsimonious and actionable framework for SME managers in decision support of e-commerce adoption.