Fig. 3: SHAP values of the 15 most important features, as determined by the XGBoost model.
From: Unveiling multiscale drivers of wind speed in Michigan using machine learning
The SHAP value represents how the model’s prediction changes in response to changes in individual features relative to the average prediction. Each number to the left of the variable name corresponds to the feature’s importance, calculated as the absolute mean of all SHAP values for that feature. This value indicates the sensitivity of monthly wind speed to variations in each feature.
