A CRITIC-Weighted Fuzzy-TOPSIS and Machine Learning Framework for Evaluating Corporate Governance Efficiency: Evidence from Construction Materials Manufacturing in Uzbekistan
Year: 2026
Type: preprint
Abstract: We develop a hybrid framework integrating CRITIC objective weighting, Fuzzy-TOPSIS governance ranking, and machine learning to evaluate corporate governance efficiency in Uzbekistan's construction materials sector. Twenty domain experts rated eight manufacturers on ten governance indicators using a 7-point Likert scale, yielding 160 observations. CRITIC assigned the highest weights to strategic planning (0.130) and risk management (0.115). Four ML models---Random Forest, XGBoost, SVR, and ANN---trained under leave-one-enterprise-out cross-validation to predict actual Return on Assets from governance indicators. The governance ranking correlated negatively with financial performance (Spearman ρ = -0.24$), revealing that expert perceptions bear little relation to profitability. SHAP analysis showed that stakeholder relations---employee and government relations---dominated the association with ROA, while structural governance dimensions contributed least. The findings document a governance paradox: the dimensions that best differentiate enterprises in expert perception are not those most closely tied to profit, and relational governance outweighs structural governance in its association with financial performance.
Source: Research Square
Institutions: Fergana State Technical University, Kokand University
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7210285402