Credit risk assessment plays a major role in the banks and financial institutions to prevent counterparty risk failure. One of the primary capabilities of a robust risk management system must be detecting the risks earlier, though many of the bank systems today lack this key capability which leads to further losses (MGI, 2017). In searching for an improved methodology to detect such credit risk and increasing the lacking capabilities earlier, a comparative analysis between Deep Neural Network (DNN) and machine learning techniques such as Support Vector Machines (SVM), K-Nearest Neighbours (KNN) and Artificial Neural Network (ANN) were conducted. The Deep Neural Network used in this study consists of six layers of neurons. Further, sampling techniques such as SMOTE, SVM-SMOTE, RUS, and All-KNN to make the imbalanced dataset a balanced one were also applied. Using supervised learning techniques, the proposed DNN model was able to achieve an accuracy of 82.18% with a ROC score of 0.706 using the RUS sampling technique. The All KNN sampling technique was capable of achieving the maximum true positives in two different models. Using the proposed approach, banks and credit check institutions can help prevent major losses occurring due to counterparty risk failure.
