DIGITAL-INTELLIGENT PENETRATING SUPERVISION AND DATA ASSET AUDIT RISK EARLY WARNING: AN EMPIRICAL STUDY ON LISTED ELECTRONIC INFORMATION FIRMS IN SICHUAN PROVINCE

Abstract

The digital economy has elevated data assets to a strategic corporate resource, yet their unique characteristics — non-rivalry, value volatility, and valuation ambiguity — introduce unprecedented audit risks. This study develops a novel theoretical framework integrating Digital-Intelligent Penetrating Supervision (DIPS) theory with audit risk assessment to construct a data asset audit risk early warning system for listed electronic information firms in Sichuan Province, China. Using panel data from 38 listed electronic information companies in Sichuan from 2023 to 2025, we propose a hybrid machine learning model combining XGBoost with SHAP (SHapley Additive exPlanations) interpretability to identify and predict data asset audit risk factors. Our empirical results reveal that (1) data asset valuation uncertainty, regulatory compliance intensity, and internal data governance maturity are the three most significant predictors of audit risk; (2) the proposed XGBoost-SHAP model achieves an AUC of 0.937, outperforming traditional logistic regression (0.781) and random forest (0.892) benchmarks; (3) firms with higher digital supervision maturity exhibit significantly lower data asset audit risk, with a marginal effect of -0.183. This study contributes to the audit literature by operationalizing the DIPS framework in the context of data asset auditing and providing a practically deployable early warning tool. Our findings have direct implications for audit practitioners, regulators, and listed companies navigating the complex landscape of data asset assurance in emerging economies.