Research Article
Predictive Modeling Approaches for Payroll Issuers
@INPROCEEDINGS{10.4108/eai.27-2-2017.152275, author={H.A. P\^{e}rez and J.A. Marmolejo and J. Velasco and J.G. Fuentes}, title={Predictive Modeling Approaches for Payroll Issuers}, proceedings={First EAI International Conference on Computer Science and Engineering}, publisher={EAI}, proceedings_a={COMPSE}, year={2017}, month={2}, keywords={Credit Scoring Logistic Regression Decision Tree Articial Neural Networks (ANNs) Ensemble Models}, doi={10.4108/eai.27-2-2017.152275} }
- H.A. Pérez
J.A. Marmolejo
J. Velasco
J.G. Fuentes
Year: 2017
Predictive Modeling Approaches for Payroll Issuers
COMPSE
EAI
DOI: 10.4108/eai.27-2-2017.152275
Abstract
Nowadays, in most banks, vast amounts of data are available in order to make business decisions and enhance the institution‘s know-how. The present study refers to transactional data systems used by companies that manage payroll outsourced services. We propose two practical approaches for analyzing this type information. One approach consists of testing traditional techniques for predictive modeling and, the other of building a credit score card using a credit scoring methodology. Several experiments were executed using specialized software in order to obtain the best credit score model for payroll issuers. Experimental results show that for most cases, decisions tree models are better than both logistic regression models and ensemble models. In one approach, we also show how the Quantile Grouping Method gives the lowest missclassication rate.