Scholarly record
ENSEMBLE CLASSIFIERS APPROACH FOR CONSUMER CREDIT SCORING
Abstract
One of basic banking activities is granting loans to both retail and corporate clients. This is obviously related to the potential risk of customers failure to settle liabilities by the agreed dates. For these reasons, banks often use very advanced techniques and tools to predict the creditworthiness of borrowers. In banks, there are special teams dedicated to this purpose dealing with broadly understood credit risk management. To assess the creditworthiness of retail clients, a wide range of analytical and statistical methods is used, including the ones based on the so-called artificial intelligence. The most frequently applied are: logit models, discriminant analysis models, decision models based on decision trees, neural networks (NN), support vector machines (SVM), etc. In recent years, ensemble classification methods aimed at maximizing classification efficiency in forecasting the event of non-payment of the loan within the set time by clients have been greatly developed. For this reason, ensemble classifiers (group classifiers) are a very good tool for applications in practical credit banking activities. This paper presents an analysis of the possibilities of using various variants of ensemble classifiers for assessing the credit risk of individual banking clients. The analysis will include three ensemble classifiers techniques: bagging, boosting and stacking, as well as various variants of individual (base) classifiers included in ensemble classifiers. Classification efficiency of best ensemble classifiers was examined in relation to standard credit scoring techniques and methods. Consumer credit risk assessment analyses made use of several datasets for bank retail clients, which are generally available in statistical repositories, as well as consumer credit data for one of the Polish credit institutions.
Publication Impact Profile
- Captures
- Mendeley - Readers: 2
Publication details
ReferencesPending
Structured references will appear here after the reference import pass. The count is preserved now so the scholarly record is not incomplete.
View or Download full articleAccess options
SWS access login
Login as SWS Scientific CommitteeLogin as SWS Scientific PartnerLogin as SWS AuthorAuthors and approved SWS contributors will read and export their own linked papers after identity matching by SWS profile, email and SGEM GlobalID.
For librarian assistance: [email protected]
Purchase Instant Access
- Article can be downloaded after successful payment.
- Article may be used according to SWS library access terms.
- Article cannot be redistributed.
