A Study on the Interpretability of Collegiate Panel Judgment Prediction Based on Concept Tree XGBoost

Currently, the majority of artificial intelligence algorithms and models employed by the Institute of Computational Law exhibit "black box" characteristics and insufficient interpretability. This field is in urgent need of models that possess both high performance and high interpretability. Therefore, this paper focuses on the prediction of collegiate panel judgments, utilizing the Concept Tree XGBoost algorithm to construct a prediction model. Furthermore, the SHAP interpretation method is integrated into this model to quantify and attribute the judgment outcomes. The experimental results demonstrate that this model can achieve favorable prediction outcomes and, to a certain extent, meet the interpretability requirements of collegiate panel judgment models in practical applications.