Fig. 4
From: An interpretable machine learning-assisted diagnostic model for Kawasaki disease in children
Local model explanation. (a,b) SHAP Waterfall plot. (c,d) SHAP Force plot. The plot a and c represent a patient with KD diagnosis to be predicted as KD, b and d represent a patient with OD diagnosis to be predicted as OD. The vertical-axis shows each feature and its actual value, and the horizontal-axis represent the Shapely Value. The arrows illustrate the influence of each feature on the model’s decision-making process, while the color represent whether the feature decreased (blue) or increased (red) the risk of KD. The combined effects of all features provided the final SHAP value \(\:f\left(x\right)\:\:\), which corresponded to the prediction score, serves as the baseline for SHAP, with its value being the mean of the model’s predicted score. The \(\:E\left[f\left(x\right)\right]\) serves as the baseline for shapely values, with its value being the mean of all the model’s prediction score.
