Automate algorithm for self-adjusting selections and thresholds in a credit scoring process.

We are looking to find a solution to a multi-stage optimisation problem where all 3 stages define a separate optimisation problem on it’s own which is, however, dependent on the other two. The 3 objectives are:

• Select the most appropriate set of metrics to be included in the scoring process;

• Allocated the most accurate weights to each of these metrics so that if the respective condition for trend or level is triggered, the weight score is added to the individual’s scorecard;

• Based on all final individual scores, find the most accurate classification thresholds’ adjustments to the previous time point of assessment.

The resulting multi-problem optimised solution can be applied in various industrial fields, such as credit scoring process.