Which algorithm is best suited if a client needs full explainability of the machine learning model?