Identifying and predicting key industry–education integration projects <i>via</i> explainable artificial intelligence (XAI) based machine learning models

Background Identifying priority projects in industry–education integration is critical for optimizing policy support and resource allocation. While prior studies have primarily relied on qualitative assessments, there remains a lack of interpretable, data-driven approaches for large-scale priority project evaluation. This study focuses on projects related to the Management Information Systems (MIS) discipline, enabling a discipline-specific analysis. Methods We developed a predictive framework by benchmarking multiple algorithms (including Logistic Regression, Support Vector Machine (SVM), Random Forest, and eXtreme Gradient Boosting (XGBoost)) on structured project-level data. The optimal tree-based ensemble model was integrated with SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) to quantify feature importance at both global and local levels. The dataset included multiple project- and enterprise-related indicators such as equipment investment, firm age, and prior project experience. Model robustness was evaluated and optimized through GridSearchCV with five-fold cross-validation. Results The optimized XGBoost model achieved a Receiver Operating Characteristic Area Under the Curve (ROC-AUC) of 0.858, demonstrating strong discriminative ability while successfully mitigating overfitting risks. Both SHAP and LIME analyses revealed that equipment investment, firm age, and project count were the top predictors. Higher equipment investment strongly contributed to positive classifications, while extremely low or high firm age often had negative impacts. Local explanations explicitly quantified the decision thresholds, reflecting heterogeneous project profiles and trade-off mechanisms influencing prediction outcomes. Conclusions The proposed framework not only delivers high predictive accuracy but also provides interpretable insights for policy formulation and project application strategy optimization. These findings demonstrate that combining ensemble learning with explainable artificial intelligence (XAI) can enhance the transparency and effectiveness of priority project identification, with strong potential for adaptation to other resource allocation and project selection tasks.