Azerbaijan’s Azericard processing center has developed a model that would use bank-card transactions to assess credit risk among people without established credit histories.
About 43% of Azerbaijan’s population has a credit history, Azericard Deputy Director General Zakir Ahmadov said at an event held as part of a fintech trade delegation program in Baku.
The model was developed jointly with Mastercard using Azericard data and is ready to be presented publicly, Ahmadov said, according to Report.az.
Model Narrows 1,350 Indicators to 56 Factors
Developers reviewed about 1,350 indicators and selected 56 main factors that may affect a borrower’s score. These include income and spending, previous late payments, the number of active loans, age, gender, account balances and patterns in other transactions.
The system uses the SHAP method to explain how individual inputs contributed to classifying a customer as a lower- or higher-risk borrower.
Ahmadov said the analysis also found that transactions made at night were associated with a greater likelihood of late payments. The finding is an observation from the model’s development and does not establish a rule for every customer.
A timetable for deploying the system, the banks that may participate and the rules governing the use of card data have not yet been announced.
Source: Vesti Baku, citing Report.az.