Credit Scoring and Credit Control Conference 2023
Creditinfo Group participated in this year’s XVIII Credit Scoring and Credit Control Conference which took place in Edinburgh on 30th August – 1st September 2023. This year’s conference focused on current industry issues and the latest credit risk analytics research findings.
Dmitry Borodin, Head of Decision Analytics Creditinfo Group, and Guilhem Poucin, Senior Risk Analyst Creditinfo Group, presented on “A Winding Road to Credit – Access to Finance for Underbanked Populations”. During their session, they demonstrated how their research examined the issue of access to finance for underbanked populations and identified actionable high impact score-driven solutions:
1️) Using Creditinfo’s global credit bureau footprint, they have explored financial inclusion of the identified populations across various territories, identifying common challenges.
2️) They then introduced a framework to address the most common challenges that limit access to credit for underbanked populations. The framework is to a large extent driven by careful utilization of alternative predictive characteristics such as payment histories on mobile or small loans, telecom and utility data, data on associated parties, and self-reported information. Furthermore, they advocated for the application of thoughtful and unbiased model development methodologies ultimately helping to enhance financial inclusion and expand access to finance for underbanked populations.
Johann Haraldsson, Data Scientist, Creditinfo Group, together with Gunnar Gunnarsson – Executive Director and Analytics and Consultant – Creditinfo Group, presented on “Chunking – A Practical Approach to Manage Consent-based Data Sources in Credit Scoring”. In this session, they presented “chunking” – a methodology used for training a credit score which is designed to handle the fact that different data sources apply to different people at different times. This applies especially to data sources where individuals must opt-in and can opt-out whenever they want. This is to ensure to simplicity, fairness and consistency through time is maintained, while losing as little statistical power as possible.
Kjartan Palsson, Data Scientist, Creditinfo Group, spoke on “Networks- A Practical Graph-Based Approach to Corporate Default Modelling”. During his session, he focused on the network of company connections, specifically how information can be incorporated from neighboring companies to enhance the company credit score. He highlighted on the following:
1) How companies can be connected through various channels, such as ownership structures, common senior staff roles or supply chain relationships. He presented a practical approach to categorizing these channels into “edge strength”, where high strength indicates that defaults are likely to spread to neighboring companies.
2) Ways to use this information to create features which are highly predictive of defaults, particularly for newly founded companies with limited financial data.
Visit www.creditinfo.com for more information.
