Local Importance Gradient (LIG) is a advanced statistical method for explainability of machine learning classification and regression algorithms. LIG quantifies the importance of features in terms of classification probability and classification change.