Methods and applications in credit scoring

Nowadays, credit cards become the most important product in developing finance sector. Onbehalf of banks, to issue or not to issue credit cards to clients should be decided after carefullyinvestigating their backgrounds. By increasing credit card demands, the evaluation of credit cardapplications become more complex system. Because credit cards specialist consider differentcriteria, the decisions can be subjective. So, the statistical and non-statistical methods are used toboth respond increasing credit card applications at right time and make objective decisions bygetting rid of subjectivity.In this work, good-bad clients are tried to be separated by regarding thirteen variables belongs tocredit card clients of a specific bank and the results of application are compared between eachother.The statistical methods used to in this work are Discriminant Analysis, Logistic Regression,Cluster Analysis and Classification Tree, and the only non-statistical method is Neural Networks.The applications of those methods are focused on without given so much details in theory.Moreover, the subjects, Lineer Programming and Integer Programming, are theoreticallymentioned in this study without giving their applications. The implementations of thosetechniques are left for further researches.Whereas there are some advantages and disadvantages of implementing those methods, theestimation rate of logistic regression is the highest among them in accordance with used data sets.Because MONTHLY NET INCOME is common variables in all models regarding to related dataset, those are observed as the most effective varibles in the credit scoring in this work.Keywords: Credit Scoring, Logistic Regression, Discriminant Analysis, Cluster Analysis,Nearest Neighbour, Classification Trees, Neural Networks.