CREDIT SCORING is a kind of binary classification problem that contains important information for manager to make a decision in particularly in banking authorities. Obtained scores provide a practical CREDIT decision for a loan officer to classify clients to reject or accept for payment loan. For this sake, in this paper a data envelopment analysis- discriminant analysis (DEA-DA) approach is used for reclassifying client to reject or accept class for case of real data sets of an Iranian bank branch. For this reason, two DEA models are solved. Also, the reject and accept frontiers and overlapping region among two frontiers are obtained. Then a goal programming problem is solved for finding coefficients of the discriminant hyper-plane. The results are obtained from the samples are kept from the main dataset, clarify that the classified hyper-plane obtained from the used method provides an almost profitable classification for payment loan.

CREDIT SCORING is becoming one of the main topics in the banking field. Lending decisions are usually represented as a set of classification tasks in consumer CREDIT markets. In this paper, we have applied a recently introduced rule generator classifier called CORER [1] (Colonial cOmpetitive Rule-based classifiER) to improve the accuracy of CREDIT SCORING classification task. The proposed classifier works based on Colonial Competitive Algorithm (CCA).In order to approve the CORER capability in the field of CREDIT SCORING, Australian CREDIT real dataset from UCI machine learning repository has been used. To evaluate our classifier, we compared our results with other related well-known classification methods, namely C4.5, Artificial Neural Network, SVM, Linear Regression and Naïve Bayes. Our findings indicate superiority of CORER due to better performance in the CREDIT SCORING field. The results also lead us to believe that CORER may have accurate outcome in other applications of banking.

CREDIT decisions are extremely vital for any type of financial institution because it can stimulate huge financial losses generated from defaulters. CREDIT SCORING models are decision support systems that take a set of predictor variables as input and provide a score as output and CREDITors use these models to justify who will get CREDIT and who will not. Many different CREDIT SCORING models have been developed by the banks and researchers in order to solve the classification problems (i.e. distinguishing the good CREDIT customers from the bad ones). Almost all these methods categorize the customers into two groups: the Good CREDITs and the Bad CREDITs. But regarding to the rapid growth in the number of CREDIT applicants and also the intense competition between financial institutions, developing the models which are able to classify CREDIT applicants into more groups (e.g. 6 or more), seems to be necessary. The purpose of this study is to propose an ANN- based algorithm which is capable of classifying the customers into 6 levels, regarding to their trust values. Till now, almost all of the studies in CREDIT SCORING are trying to improve the accuracy rate of the proposed algorithms and this is the first time that trust's concept is used in CREDIT SCORING domain. On the other hand, categorizing customers into more groups, will lead to make fast, easy, certain and fair CREDIT lending decisions.

CREDIT SCORING is an important topic, and banks collect different data from their loan applicant to make an appropriate and correct decision. Rule bases are of more attention in CREDIT decision making because of their ability to explicitly distinguish between good and bad applicants. The CREDIT SCORING datasets are usually imbalanced. This is mainly because the number of good applicants in a portfolio of loan is usually much higher than the number of loans that default. This paper use previous applied rule bases in CREDIT SCORING, including RIPPER, OneR, Decision table, PART and C4.5 to study the reliability and results of sampling on its own dataset.A real database of one of an Iranian export development bank is used and, imbalanced data issues are investigated by randomly Oversampling the minority class of defaulters, and three times under sampling of majority of non-defaulters class. The performance criterion chosen to measure the reliability of rule extractors is the area under the receiver operating characteristic curve (AUC), accuracy and number of rules. Friedman’s statistic is used to test for significance differences between techniques and datasets. The results from study show that PART is better and good and bad samples of data affect its results less.