Probability calibration: A tool to mitigate the risk of your Machine learning model.

The banking industry has been applying machine learning to enhance loan approval process, default risk assessments, and fraud detection, for example. Making accurate decisions in these areas is important for effective risk management and cost control, which can translate to significant value. While traditional approaches and machine learning methods can be used to determine creditworthiness and predict loan defaults, predictions are based on historical patterns and therefore come with uncertainty. To control uncertainty, banks need to effectively quantify and manage the risk to prevent unexpected losses. FICO score is an example of a widely used metric to measure creditworthiness of customers [<a href="https://www.investopedia.com/terms/f/ficoscore.asp" rel="noopener ugc nofollow" target="_blank">Adam</a>]. However, banks must still meet the minimum total capital requirements according to Risk-Based Capital Requirement [<a href="https://www.bis.org/basel_framework/standard/RBC.htm" rel="noopener ugc nofollow" target="_blank">Basel</a>] to cover potential losses, which ties the capital. <a href="https://medium.com/scb-datax/probability-calibration-a-tool-to-improve-your-fairness-of-your-machine-learning-model-faba02cc9dca">Read More</a>