Explanation of Machine Learning Models Using Shapley Additive Explanation and Application for Real Data in Hospital

TL;DR

This paper uses SHAP to interpret hospital data-driven machine learning models, introducing new feature importance metrics and feature packing to improve interpretability and demonstrate clinical relevance.

Contribution

It proposes novel SHAP-based interpretability techniques, including a new feature importance metric and feature packing, for better understanding of hospital data models.

Findings

SHAP provides clear explanations of model predictions.

Feature packing simplifies model interpretation without retraining.

A/G ratio identified as a key prognostic factor for cerebral infarction.

Abstract

When using machine learning techniques in decision-making processes, the interpretability of the models is important. In the present paper, we adopted the Shapley additive explanation (SHAP), which is based on fair profit allocation among many stakeholders depending on their contribution, for interpreting a gradient-boosting decision tree model using hospital data. For better interpretability, we propose two novel techniques as follows: (1) a new metric of feature importance using SHAP and (2) a technique termed feature packing, which packs multiple similar features into one grouped feature to allow an easier understanding of the model without reconstruction of the model. We then compared the explanation results between the SHAP framework and existing methods. In addition, we showed how the A/G ratio works as an important prognostic factor for cerebral infarction using our hospital data and proposed techniques.