SHAP (SHapley Additive exPlanations) is a game theoretic approach to explain the output of any machine learning model. It connects optimal credit allocation with local explanations using the classic Shapley values from game theory and their related extensions.
The main idea of SHAP rests upon the following equation where it captures the average marginal contribution of a feature value (player) across all possible coalitions (society).
where n is the total number of players and the sum extends over all subsets S of N not containing player i. The formula can be interpreted as follows: imagine the coalition being formed one actor at a time, with each actor demanding their contribution as a fair compensation, and then for each actor take the average of this contribution over the possible different permutations in which the coalition can be formed.
Reference: Shapley Values
The enviornment is as follows:
- python 3.6.x
- windows 10
- scikit-learn
- shap
The dataset used in this notebook is taken from the Kaggle-titanic competition. The main objective is to predict wheather a passenger will survive.
Features Summary Plot
Features Summary Bar
Explained by Sex
Explained by Pclass
Explained by Age
