Interpreting Black-box Machine Learning Models for High Dimensional Datasets