Application Score

Assessment of default probability based on the borrower’s application data.

This decision is particularly useful for assessing the risk of issuing a credit to borrowers whose credit history is either uninformative or absent.

Contents
Identification

Borrower’s current ID data and the history of changes made to it.

Scoring

A value from 0 to 1 that determines the probability of late payments of over 100 UAH that are at least 90 days past due within a 6-month period on a credit being considered.

Application Scoring includes:

Model development;

API development and implementation;

10,000 free reports for testing;

Monthly monitoring of the quality of model distribution;

Model calibration during one year after implementation, in case of distribution changes.

Model
Target function
Late payment of over 100 UAH at least 90 days past due within a 12-month period
Predictors
Credit history characteristics and partner application data
Method
Gradient boosting
0.79
ROC-AUC
0.49
KS
Default probability based on scoring
Working with the scoring

To build an individual scoring, the following is required:

Sample for model training. The required number of records is 10,000 or more;

Sample markup into «good» and «bad»;

Description of the business process for scoring application;

Description of data for model training.

Details and terms of development are negotiated individually.

XML
XML format is used to work with scoring.
General structure of interaction:
Http Method
POST
Request URL
Test URL
Request
Response

Do you want to become a partner of the Ukrainian Bureau of Credit Histories?

Apply for cooperation