Predictive Model for the Evaluation of Credit Risk in Banking Entities Based on Machine Learning

In this paper, we propose a technology model of predictive analysis based on machine learning for the evaluation of credit risk. The model allows predicting the credit risk of a person based on the information held by an institution or non-traditional sources when deciding whether to grant a loan. In this context, the financial situation of borrowers and financial institutions is compromised. The complexity of this problem can be simplified using new technologies such as Machine Learning in a Cloud Computing platform. Azure was used as a tool to validate the technological model of predictive analysis and determine the credit risk of a client. The proposed model used the Two-Class Boosted Decision Tree algorithm that gave us a greater AUC of 93% accuracy, this indicator was taken as having greater repercussion in the proof of concept developed because it is wanted to predict more urgently the number of possible applicants who do not comply with the payment of debits.