The research aims to identify critical drivers of enterprise digital transformation by integrating corporate culture (meso-level) and executive characteristics (micro-level), providing actionable insights for optimizing cultural systems, leadership strategies, and stage-specific transformation roadmaps.
This study employs machine learning (ML) algorithms (Random Forest, LightGBM, XGBoost) and SHAP analysis to examine how corporate culture and executive traits influence digital transformation. Using 2017–2023 data from Chinese A-share firms, it quantifies cultural dimensions via Word2Vec and constructs a four-stage digital transformation evaluation system. Ensemble models address multicollinearity and nonlinear relationships, while SHAP interprets feature contributions across transformation stages.
Quality culture, innovation culture, and internationalization culture are core drivers. Executive tenure and shareholding ratios significantly influence transformation outcomes. SHAP reveals nonlinear interactions, with partial cultural/leadership demand variations across digital transformation stages.
Enterprises should prioritize quality-centric cultural systems, balance executive tenure incentives, and align leadership shareholding ratios with transformation goals. Stage-specific strategies should leverage cultural strengths (e.g. international culture in advanced stages) and mitigate overemphasis on ethics hindering innovation.
Findings support sustainable digital economy growth by clarifying organizational governance mechanisms. Policymakers can refine digital infrastructure policies, while firms gain frameworks to overcome “transformation paralysis” and align stakeholder interests.
First, it utilizes ML ensemble algorithms to make overall predictions of the degree of enterprise digital transformation, systematically examining the predictive performance and differences of corporate culture and multi-dimensional executive characteristics on enterprise digital transformation. Second, this paper quantifies the six-dimensional characteristics of corporate culture through the ML method Word2Vec, broadening the research scope of corporate culture factors and their quantitative indicators. Third, by constructing a four-stage evaluation system for digital transformation, this paper uses ML ensemble algorithms combined with the SHAP tool to deconstruct the complex influencing factor paths of digital transformation.
