A Knowledge Distillation-Based Approach to Enhance Transparency of Classifier Models