What is SHAP? Why do we use it in One AI and on One AI Storyboards?

Answer:

SHAP is a powerful mathematical method to explain the predictions of machine learning models. SHAP provides the contribution of each feature to each individual prediction. It is used on storyboards to provide a numerical way to model the data and to increase transparency and interpretability of your model.

Example of SHAP Values modeled on a Storyboard

By using advanced visualization techniques on SHAP data, a “picture” of the model can be formed. An example of this is the SHAP “beeswarm” chart.

The main idea behind SHAP is rooted in cooperative game theory, specifically the Shapley value concept, which assigns a value to each player (feature) in a coalition (subset of features) based on their marginal contribution to the overall outcome.

• SHAP values quantify the impact of each feature on the model's prediction for a particular instance.
• Positive SHAP values indicate features that contribute to increasing the prediction, while negative values indicate features that decrease the prediction.
• SHAP values offer a local explanation for individual predictions, providing transparency and understanding into why a model made a specific prediction for a given input.

Please note that SHAP is not enabled on machine learning models by default which can be accomplished in the “Global” augmentation configuration settings.  See this external SHAP explanation for more information about SHAP. 

Generate SHAP Values from the Global Settings of the Machine Learning model