A multimodel optimization and feature contribution analysis for evaluating outsourced suppliers of equipment manufacturing enterprises

The deepening advancement of the integration strategy has imposed dual challenges on the equipment manufacturing enterprises procurement system: supplier diversification and scale expansion. Traditional supplier management models, constrained by information barriers and institutional limitations, struggle to meet the demand for precise evaluation. To address persistent pain points in the defense supply chain — including fraudulent qualification claims, unstable fulfillment quality, and static evaluation metrics — this study innovatively proposes a supplier assessment methodology integrating multi-model optimization and feature contribution analysis. First, by deconstructing the production modes (i.e., supply chain structures) of multi-variant equipment manufacturing products, we embed upstream and downstream parameters into the evaluation framework, thereby establishing a comprehensive indicator system for outsourcing supplier assessment in equipment manufacturing production. Subsequently, we develop an intelligent evaluation approach that synergizes multi-model selection and feature contribution analysis. This methodology holistically incorporates diverse machine learning models capable of capturing nonlinear correlations, enabling automatic selection of the optimal predictive model. It further performs contribution analysis on evaluation outcomes to iteratively refine and optimize the indicator system. The proposed method provides a technical pathway to resolve the coupling dilemma between administrative directives and market mechanisms, empowers digital transformation of defense supply chains, and enriches the theoretical framework of integration — demonstrating both academic significance and practical implications.