Uso de una red neuronal para determinar el desempeño de una organización del sector bancario

Abstract

In Ecuador, there are a vast quantity of credit unions, 852 in total, which are within 5 different segments by their assets, but only 35 of 58 credit unions of segments 1 and 2 present risk rating records, and this analysis is almost non-existent for segment 3, 4 and 5, besides about 66% of credits in Ecuador come from credit unions. This study pretends to create an artificial neural network (ARN) with an acceptable accuracy of classification for this type of companies using their financial ratios in order to be used as a helper to support financial decisions. The ARN was created by data obtained from SEPS (Superintendencia de economía publica y solidaria) for credit unions with a high assets value and using a record of risk rating since 2015 to 2017; the data used to this study is public access. Then, an ARN was created using a hierarchy classification level, it presented an accuracy of 83.19%, and another ARN was created using direct classification with 82.65% accuracy. Finally, due to the first ARN presents slightly better results, this tool was defined as the best option of this study and it can be used as a form of support for classification of Union credits, moreover this tool presents a great potential to increase its accuracy and become a main tool for classification and prognostics.