The increase of governmental functions in modern society leads to the increasing number of departments, while the lack of interdepartmental coordination causes administrative inefficiency. This study addresses the problem of multi-departmental coordination in government administration and uses dynamic programming and XGBoost algorithm to construct an assessment model to explore the key factors affecting the coordination of government departments and their role mechanisms. Based on the data from the Spike Good Office platform in Guangdong Province, the study analyzes the effects of four key variables, namely cognitive bias, imbalance of power and responsibility, lack of institutional protection and government support, on the coordination effect. The results show that the XGBoost model has a high-precision prediction ability with an accuracy of 98.13%, an AUC value of 96.37%, and a KS value of 85.24%, which can effectively differentiate the advantages and disadvantages of multi-sectoral coordination effects; the SHAP analysis reveals that cognitive bias has the most significant effect, with an average SHAP value of 0.102458, followed by lack of safeguards and imbalance of power and responsibility, with an average SHAP values of 0.033274 and 0.01894, respectively; partial dependency plots further reveal that the effect of multisectoral coordination is more stable when the cognitive bias is in the 3.0%-3.4% range, the imbalance of power and responsibility is in the 7-10 years range, and government support is greater than 13.5%. The study is of practical guidance for improving the government multisectoral coordination mechanism.
