Intelligent predictive model applying Data Mining strategies for a credit evaluation of a commercial company (#1148)

Authors

Castañeda Rojas, Grecia

Canales Carreño, Schleiffer

Ovalle, Christian

Rabanal Chávez, Erick Humberto

Abstract

This scientific article presents a predictive model developed by the Orange software, to evaluate the credit capacity of clients, through their transactions of a commercial company, with the objective of preventing delinquency and lack of cash flow. The model is guided by the SEMMA methodology and uses neural network, logistic regression and decision tree algorithms, and its accuracy was measured by performance indicators. The results showed that the decision tree algorithm achieved an accuracy of 99%, which demonstrates the efficiency of the model and the prediction if the client will comply with the payment. In addition, a significant decrease in the time required to assess the creditworthiness of clients was identified after the implementation of the intelligent predictive model. Before the model, 9 human operations were required to assess credit, while after the model it was reduced to only 6 human operations. This translated into a reduction in operating time of 33.33%. In addition, the implementation of the predictive model also made it possible to significantly reduce the time required to complete the first workflow. Before the model, the collection process could take from 60 to 240 days, but after the implementation of the model, the collection time was reduced to only 60 days. In addition, the implementation of the model was also able to completely eliminate delinquent customers, indicating a significant improvement in the company's credit risk management and productivity improvement.