The efficiency investigation of deposit banks in turkey with data envelopment analysis (DEA), C5.0 and CART algorithm: 2009-2017 period

Data mining has developed the decision-making function with the data that has become almost impossible against the increasing amount of data from the past to the present and has developed into the literature as a discipline that provides the opportunity to work with large data. In this study, the efficiency scores of the banks were obtained by using Data Envelopment Analysis, and then one of the four data mining models developed for this classification was determined as the most suitable model based on accuracy rates. The difference between the four models is based on the two different data sets they use. The basic data set used in the study consisted of 30 financial ratios and efficiency scores of 198 units. First of the data sets was reached applying the Mann-Whitney U test to the basic data set. The first two models were obtained applying C5.0 and CART algorithms to this data set. The second data set consisted of the financial ratios obtained as a result of the Binary Logistic Regression analysis applied to the basic data set. The last 2 models were created with C5.0 and CART algorithms applied to this set. As a result, the highest performance model was obtained by CART algorithm applied to Data Set-1. Financial ratios that make the highest contribution to the efficiency score were (35,2%) total deposits/total assets, (19,5%) (equity-fixed assets)/total assets, (10,3%) TL deposits/total deposits, (9,7%) non-performing loan (gross)/total loans and receivables.