Explanation of Machine Learning Models of Colon Cancer Using SHAP Considering Interaction Effects
This work addresses interpretability for medical decision-making, but it is incremental as it adapts an existing method to a specific domain.
The authors tackled the problem of distinguishing main and interaction effects in SHAP-based interpretability of machine learning models, applying the Shapley-Taylor index to colon cancer data from 29,080 patients to identify risk factor combinations.
When using machine learning techniques in decision-making processes, the interpretability of the models is important. Shapley additive explanation (SHAP) is one of the most promising interpretation methods for machine learning models. Interaction effects occur when the effect of one variable depends on the value of another variable. Even if each variable has little effect on the outcome, its combination can have an unexpectedly large impact on the outcome. Understanding interactions is important for understanding machine learning models; however, naive SHAP analysis cannot distinguish between the main effect and interaction effects. In this paper, we introduce the Shapley-Taylor index as an interpretation method for machine learning models using SHAP considering interaction effects. We apply the method to the cancer cohort data of Kyushu University Hospital (N=29,080) to analyze what combination of factors contributes to the risk of colon cancer.