Topological conceptual framework for multicriteria inventory classification criteria selection: process dynamics in a steel manufacturing case
针对多准则库存分类中准则选择易受主观偏差影响的问题，提出一个拓扑框架来诊断上游过程扭曲，并通过钢铁制造案例验证了专家准则集存在统计显著的结构异常。
Criteria selection remains a critical vulnerability in Multicriteria Inventory Classification (MCIC), often compromising model validity through subjective bias. This study addresses this deficit by introducing a topological framework to diagnose upstream process distortions. We map the criteria selection of a steel manufacturer to topological constructs, revealing specific pathologies: functional disconnection, boundary instability, and latent disagreement masking as consensus. To validate these insights, we employ persistent homology and Monte Carlo simulation, benchmarking the expert-derived set against 2,000 null models generated from an organisational text corpus. The analysis identifies the expert set as a statistically significant structural outlier (p < 0.001). Unlike natural decision spaces, the expert set exhibits anomalous topological complexity and high-persistence cycles (β1), indicating that expert judgement imposes structural cyclicity upon the criteria configuration. These findings demonstrate that cognitive and procedural factors manifest as measurable geometric distortions, establishing Computational Topology as a diagnostic tool for assessing the structural properties of criteria sets before mathematical aggregation occurs.