Bank Customer Credit Scoring by Using Fuzzy Expert System

Abstract

Abstract-Granting banking facility is one of the most important parts of the financial supplies for each bank. So this activity becomes more valuable economically and always has a degree of risk. These days several various developed Artificial Intelligent systems like Neural Network, Decision Tree, Logistic Regression Analysis, Linear Discriminant Analysis and etc, are used in the field of granting facilities that each of this system owns its advantages and disadvantages. But still studying and working are needed to improve the accuracy and performance of them. In this article among other AI methods, fuzzy expert system is selected. This system is based on data and also extracts rules by using data. Therefore the dependency to experts is omitted and interpretability of rules is obtained. Validity of these rules could be confirmed or rejected by banking affair experts.For investigating the performance of proposed system, this system and some other methods were performed on various datasets. Results show that the proposed algorithm obtained better performance among the others.

Configuración de un Control de Temperatura en un Sistema Embebido de Bajo Costo, usando Herramientas de Inteligencia Artificial y el Internet de las Cosas

“…En 1965, Lofti A. Zadeh propuso el álgebra para la operación de estos conjuntos (Zadeh, 1965), y diferentes representaciones han hecho uso de esta, como, por ejemplo, los sistemas difusos Mamdani (Mamdani & Assilian, 1975), con variables lingüísticas de entrada y salida, que poseen sistemas de inferencia basados en reglas de la forma SI antecedente, ENTONCES consecuente, cuyo fin es obtener conclusiones de las variables lingüísticas de salida, a partir de los actuales valores de las variables de entrada. En los últimos años se han desarrollado un gran número de aplicaciones principalmente en sistemas de control automático, que se construyen a partir de reglas semánticas y el conocimiento experto (Olufunke, Charles, Charles, Abraham, & Snasel, 2013) (Bazmara & Donighi, 2014).…”

The applicability of credit scoring models in emerging economies: an evidence from Jordan

“…Furthermore, ANN gives better accuracy rates than LR. Bazmara and Donighi (2014) used fuzzy rules expert systems to evaluate bank customer credit in one bank in Iran. They benchmarked their results with methodologies in other studies such as LR and ANN showing better performance.…”

Hybridized artificial neural network classifiers with a novel feature selection procedure based genetic algorithms and information complexity in credit scoring

“…In the credit scoring literature, some remarkable studies are based on discriminant analysis 3,4 ; logistic regression (LR) analysis, 5,6 decision trees (DT), 7‐9 linear and nonlinear programming, decision support systems, ANNs, SVMs 6,10‐22 ; Bayesian networks, 23,24 expert systems and hybrid approaches 1,4,6,25‐31 …”