Topic: Contributions on MCMC and Bayesian econometrics Title: Evaluating credit worthiness in a non credit culture society Authors:  Christson Adedoyin - Samford University (United States)
Mary Akinyemi - University of Lagos (Nigeria) [presenting]
Abstract: In the current world system where credit drives majority of transactions, it becomes increasingly necessary to evaluate credit worthiness. In a non credit culture society, that is a system where individuals have no credit history, evaluating the credit worthiness of an individual becomes a daunting task. We consider the key factors that would determine the credit worthiness of an individual and attempt to evaluate credit worthiness of customers in a Nigerian bank using statistical and artificial intelligence techniques such as Random Forests, Logistic Regression, Classification trees, Support vector machines, Neural networks and Linear discriminant analysis. We found that for all the models seem to correctly classify the credit worthy customers. However, Random Forest, Logistic Regression and Support Vector Machine better classify individuals who are not credit worthy correctly. Furthermore, Random Forest, Logistic Regression and Support Vector Machine have the highest success rates. Also, Random Forest, Logistic Regression and Support Vector Machine realizes less cost of misclassification. In addition, the Mean Decrease Gini indicates that gender and education are major factors that determine the credit worthiness of a customer in a Nigerian bank.