Risk optimization of small business in banking domain

Application Track:

Code:

Domain:

Proposed by:

Entity Logo:

Summary of the entity:

Yapı Kredi Teknoloji is a Turkish company that delivers outputs in machine learning, data mining, pattern recognition, artificial intelligence and natural language processing fields and develops mobile applications.

Summary of the challenge:

YKT aims to identify the customers that they are probable to be default before any delinquency occurs in one of the scheduled payments of any products.

Description of the global challenge:

Non-Performing Loan Detection (NPL) is one of mostly studied problem in commercial banking. At the beginning of each financial year/term, the banks specify their risk appetite in terms of what is the maximum tolerable amount of undischarged debts in a specific time frame. During each term, it is aimed to collect all late payments from default credits.

According to the Banking Regulation and Supervision Agency of Turkey, rates of commercial loan applications are highly increasing which makes detecting NPL a more crucial safety measure.

Being one of the top banks in Turkey, YKT has access to a lot of important data such as data in the loan applications, demographic data of the clients, financial documents of the applicant companies, credit bureau data related to the company, payment behaviour of the client, among others.

Before any delinquency occur in one of the scheduled payments of any products, YKT aims to identify the customers that they are probable to be default. Combining these with the state-of-the-art machine learning methods and experience in the field, YKT intends to predict NPL before 6 months of period and decrease the ratio of NPL.

Sub-challenges composing this experiment:

This challenge is composed of 2 sub-challenges:

• Impact of Demographic Determinants of NPL (REACH-2020-READYMADE-YKT_2.1)
• Impact of Financial Determinants of NPL (REACH-2020-READYMADE-YKT_2.2)

Expected global results:

This challenge is designed to employ Machine Learning (ML) algorithms on financial data. Especially, feature engineering techniques on time series tabular data is focused on during the development. By monitoring the existing portfolios, it is aimed that 20% of NPL portfolios can be recovered with early tracking functionalities, which is equal to € 400 000 saving per year.

IMPACT OF FINANCIAL DETERMINANTS OF NPL

Code:

REACH-2020-READYMADE-YKT_2.2

Summary of the sub-challenge:

The goal is to improve YKT model by analysing the effect of financial data of customers on the non-performing loans.

Description of the challenge:

Non-Performing Loan Detection (NPL) is one of mostly studied problem in commercial banking. At the beginning of each financial year/term, the banks specify their risk appetite in terms of what is the maximum tolerable amount of undischarged debts in a specific time frame. During each term, it is aimed to be collect all late payments from default credits.

According to the Banking Regulation and Supervision Agency of Turkey, rates of commercial loan applications are highly increasing which makes detecting NPL a more crucial safety measure.

Being one of the top banks in Turkey, YKT has access to a lot of important data such as data in the loan applications, demographic data of the clients, financial documents of the applicant companies, credit bureau data related to the company, payment behaviour of the client, among others.

Within the global YKT challenge to predict NPL, this specific sub-challenge aims to improve YKT model by analysing the effect of financial data of customers on the non-performing loans.

Data to be used:

• Small Business NPL Detection

Expected outcomes:

By monitoring the existing portfolios, it is aimed that 20% of NPL portfolios can be recovered with early tracking functionalities, which is equal to € 400 000 saving per year.