SHAP values decompose a prediction into the contribution of each feature, ensuring that the sum of the contributions equals the difference between the actual prediction and the average prediction. This method allocates the contribution in a way that is fair and consistent, making it easier to understand the influence of individual variables.
