Credit Scoring Model Based on HMM/Baum-Welch Method
| Autor: | Abdelhak Zoglat, Mohamed Ouzineb, Badreddine Benyacoub, Souad Elbernoussi |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Iterative and incremental development
050208 finance Training set Computer science business.industry 05 social sciences Economics Econometrics and Finance (miscellaneous) Machine learning computer.software_genre Computer Science Applications Task (project management) Set (abstract data type) symbols.namesake ComputingMethodologies_PATTERNRECOGNITION 0502 economics and business symbols Artificial intelligence 050207 economics Procedure approach Hidden Markov model Baum–Welch algorithm business Area under the roc curve computer |
| Zdroj: | Computational Economics. 59:1135-1154 |
| ISSN: | 1572-9974 0927-7099 |
| DOI: | 10.1007/s10614-021-10122-9 |
| Popis: | Credit scoring becomes an important task to evaluate an applicant by a banker. Many tools are available for making initial lending decisions. This paper presents a Hidden Markov Model (HMM ) for credit scoring, and uses Baum-Welch method; an iterative procedure approach; for building a set of credit scoring models. We introduce HMM/Baum-Welch model: a tool developed to explore a good accurate model for classification problems. There are two phases in this model: learned an initial model from training data using HMM, and re-estimating HMM parameters by an iterative process using Baum-Welch algorithm. The proposed model is successfully applied to a real credit problem, and the application procedure is illustrated through two data sets: German and Australian. The criteria used to evaluate the performance of different resulting models are the accuracy and AUC (area under the ROC curve). The experiment of this model, shows that, HMM with Baum-Welch approach can improve the pattern classification performance in credit scoring. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full text from SpringerLink