Credit score cards are widely used in commercial banking. As a part of risk management, personal information of the clients are gathered and analyzed to predict of probability of default and the allowance for borrowing. Commer- cial banks are capable of bridging the gap between personal information and potential risk of clients by developing algorithms such as machine learning.
Build a machine learning model to predict the credit level of a particular client so that commercial banks could distribute loans accordingly and mini- mize their risk. In order to reduce the overall risk level, the bias of the data (pre-existing, technical and emergent) should be mitigated. In this way, we could generate more accurate outcomes that precisely predict the credibility of customers. A client is supposed to be labeled as ”good” or ”bad” based on the algorithm developers’ discretion. Also, there should be mitigation methods to tame the unbalanced data in this task. So, there may be some trade off between accuracy and fairness.