Advantages and limitations of SHAP
In the previous section, we discussed the practical application of SHAP for explaining a regression model with just a few lines of code. However, since SHAP is not the only explainability framework, we should be aware of the specific advantages and disadvantages of SHAP, too.
Advantages
The following is a list of some of the advantages of SHAP:
- Local explainability: Since SHAP provides local explainability to inference data, it enables users to analyze key factors that are positively or negatively affecting the model's decision-making process. As SHAP provides local explainability, it is useful for production-level ML systems, too.
- Global explainability: Global explainability provided in SHAP helps to extract key information about the model and the training data, especially from the collective feature importance plots. I think SHAP is better than LIME for getting a global perspective on the model. SP-LIME in LIME is good for...