KakaoBank's Credit Evaluation Technology Research Published in International Academic Journal

[Asia Economy Reporter Eunju Lee] KakaoBank announced on the 29th that it has published a research paper on a credit scoring model that provides theoretical grounds for improving credit evaluation performance in a prestigious international academic journal.

According to KakaoBank, the paper titled "Two-stage credit scoring using Bayesian approach," jointly developed by the Financial Technology Research Institute (Dr. Shin Jin-ho) and the Big Data Analysis Team (Dr. Kyung Sung-hyun), has been published in a renowned international academic journal.

The research paper was published in the "Journal of Big Data," an international journal specializing in big data theory and methodology. The Journal of Big Data has an Impact Factor of 10.8, placing it among the top 5% of international journals in the field of big data theory and methodology.

The "two-stage credit scoring model" is a methodology for developing credit scoring models that can apply more explanatory variables (types of data) than existing models, enabling improved evaluation performance using a wider variety of variables while simultaneously maintaining the model's interpretability.

Most financial companies develop credit scoring models using logistic regression analysis, which assumes that explanatory variables are mutually independent, limiting the model to about 10 variables. Even if new variables are discovered, the independence constraint prevents the use of existing variables, posing a limitation on performance improvement.

Attempts to improve credit scoring model performance using machine learning and deep learning technologies are also ongoing. However, while these technologies can utilize hundreds of data points, the functional relationship between variables and outcomes is not revealed, making them difficult to apply in the financial sector where interpretability in the credit evaluation process is essential. KakaoBank's research is significant in that it proposes a methodology and theoretical basis that greatly expands the number of variables to improve performance while perfectly maintaining interpretability.

A KakaoBank representative stated, "By continuously developing credit scoring model technologies, we will provide a reasonable evaluation system for customers with limited financial history and strengthen financial inclusion."

This content was produced with the assistance of AI translation services.