A Robust Method to Measure the Global Feature Importance of Complex Prediction Models

Using SHAP to analyze technical and tactical performance in table tennis based on hybrid Catboost-TPE with four-phase evaluation theory

“…The obtained conversion equation is substituted by the derivation of the above conversion equations into the TE equation, as shown in Equation (3). The equations for calculating each phase of TEs in the new three-phase evaluation theory are obtained, as shown in Equations (22)(23)(24).…”

“…The specific processions are shown in Equations (18)(19)(20)(21)(22)(23)(24). Meanwhile, by the derivation result of the above conversion equations, the equations for calculating each phase of TEs in the new three-phase evaluation theory are obtained, as shown in Equations (22)(23)(24). The value of using the data multiple times was improved by applying two table tennis technical and tactical analysis theories through one data collection based on the four-phase evaluation theory.…”

“…Therefore, the study derived the equations for converting the SRs and URs from the four-phase evaluation theory into the new three-phase evaluation theory. The specific processions are shown in Equations (18)(19)(20)(21)(22)(23)(24). Meanwhile, by the derivation result of the above conversion equations, the equations for calculating each phase of TEs in the new three-phase evaluation theory are obtained, as shown in Equations (22)(23)(24).…”

“…Firstly, previous studies have focused too much on improving the models' performance without explaining them [20,21,22]. Artificial intelligence algorithms are known as "black boxes" due to the complexity of their internal working mechanisms and decisionmaking progress, which are far beyond human understanding [23,24,25,26,27,28]. Therefore, previous studies have used simulation ways to find the key factors in matches [20,21,22], which significantly reduced the reliability of the analytical results.…”