The Credit Scoring solution is a model-based decision system for the classification of customers as creditworthy or not. The solution assigns to each customer a probability of defaulting, translated into a score. The total score of a customer is decomposed into components, each connected to a particular characteristic (e.g. age, profession, income range, etc. for persons or analogous data for business entities). The outcome of the process is the construction of a scorecard which the organization can use to check and accept or reject applicants.
The underlying methodologies are usually based on the traditional (and well performing in such situations) logistic regression model. The best model found on a training dataset is fine-tuned on a validation dataset and finally verified on an unseen dataset.
The cut-off limits for the accept / reject decisions are agreed with the customer, according to his credit policy. The data management and the statistical analysis are implemented in SAS, R or Python.