Delinquency Analysis on Billing Data Using Tree Augmented Naive Bayesian Network

• Cássio Denner Noronha Vinhal UFG
• Gélson da Cruz Júnior UFG
• Luciana de Oliveira Berretta UFG

DOI: https://doi.org/10.5753/sbsi.2006.14754

Abstract

The present work verifies the applicability of Bayesian Bayesian Classifiers in billing databases of a power distribution company. distributor. The goal is to find patterns or profiles in certain consumption groups and groups and estimate the number of defaulters. The computational system identifies patterns in the history of each customer and projects probable behaviors. We used the Bayesian Augmented Ingénue Bayesian classifier in Bayesian classifier was used as opposed to the Naive Bayesian. The validation is done by comparison of the prediction hit rates. The conclusions indicate an adequate approach that offers subsidies for establishing effective trade policies and establish effective commercial policies and reduce defaults.

Keywords: Bad Debt, Billing Data, Tree Augmented Naïve-Bayes

References

BERRETTA, L.O. (2005), “Análise de Inadimplência em Dados de Faturamento Utilizando Rede Bayesiana Ingênua Aumentada em Árvore”, Dissertação de Mestrado, EEEC – UFG, Goiânia, Brasil.

CHOW, C., LIU, C. (1968), “Approximating Discrete Probability Distributions with Dependence Trees”. IEEE Transactions on Information Theory, vol.14-3, 462- 467, USA.

CHENG, J.; GREINER, R. (2000) “Comparing Bayesian Network Classifiers”. Alberta, CA: University of Alberta, 2000.

HAN, J.; KAMBER, M. (2000) “Data Mining, Concepts and Techniques”. USA, Morgan Kaufmann.

MELLO, L. C. (2001) “Uma revisão de abordagens genético-difusas para descoberta de conhecimento em banco de dados”. Porto Alegre, RS: Universidade Federal do Rio Grande do Sul - UFRS.

PEARL, J. (1988) “Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference”. USA, Morgan Kaufmann.

YANG, Y. (2003) “Discretization for Naïve-Bayes Learning”. [S.l.]: School of Computer Science and Software Engineering of Monash University.