Comparative Study of Two Feature Selection-Based Risk Assessment Models for Industrial Software

China dominates global manufacturing but lags in industrial software, exposing critical vulnerabilities amid rising risks like supply chain attacks and geopolitical disruptions. This study constructs a three-tier risk system with 20 metrics, including open-source shutdowns and data leaks. A novel SVM model is proposed, using hybrid kernels to capture nonlinear interactions and supply chain topologies, optimized via Bayesian methods. Evaluated on industrial datasets, SVM-ISRA outperforms AHP-FCE. While excelling in automation and dynamic adaptation, SVM-ISRA's “black-box” nature necessitates hybrid frameworks integrating expert knowledge for high-stakes compliance. The work demonstrates SVM's superiority in complex risk environments while advocating explainable AI extensions to bridge data-driven insights with hierarchical risk structures.