Objašnjivi modeli strojnog učenja za procjenu kreditnog rizika

master's thesis

• Title (english): Explainable Machine Learning Models for Credit Risk Assessment
• Author: Toma Puljak
• Mentor: Zvonko Kostanjčar (mentor)
• Committee member: Zvonko Kostanjčar (predsjednik povjerenstva)
• Committee member: Tomislav Petković (član povjerenstva)
• Committee member: Stjepan Begušić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2023-07-13, Croatia

Abstract

Zbog regulacija koje su postavljene od Europskog bankovnog tijela, svi modeli strojnog učenja koji se koriste u procjeni kreditnog rizika moraju biti u potpunosti objašnjivi. Zbog tih regulacija, financijske institucije pri procjeni kreditnog rizika koriste jednostavnije modele strojnog učenja, poput logističke regresije, koji često nemaju najveću točnost. Ovaj diplomski rad istražuje različite slojeve objašnjivosti koje treba zadovoljiti model "crne kutije", specifično XGBoost, kako bi bio u skladu s trenutnom regulativom i predlaže izračun doprinosa značajki promjeni PD-a i migraciji kategorije rizika. Obje su implementirane metode zadovoljile trenutne propise i financijske institucije bi ih mogle koristiti za implementaciju XGBoosta za procjenu kreditnog rizika.

Abstract (english)

Because of regulations posed by the European Banking Authority, all machine learning models used in credit risk assessment need to be fully explainable. This made financial institutions resort to using simple machine learning models, such as logistic regression, when assessing credit risk, which often do not have highest accuracies. This thesis explores different explainability layers that need to be satisfied by a "black-box" model, specifically XGBoost, to comply with current regulation and proposes the calculation feature contributions to PD change and risk category migrations. Both methods proved to satisfy current regulations and could be used by financial institutions to deploy XGBoost for credit risk assessment.

• Keywords: Objašnjivo strojno učenje; XGBoost; Procjena kreditnog rizika
• Keywords (english): Explainable machine learning; XGBoost; Credit risk assessment
• Language: english
• URN:NBN: urn:nbn:hr:168:076269
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-10-11 10:15:33