A New Predictive Machine-learning Approach for Detecting Creditworthiness of Borrowers

We present in this paper a new approach for predicting creditworthiness of borrowers that we call “Method of splitting the learning set into two region”. The aim of this approach consists on the construction of two regions from a learning set, the first called "Solvency Region" that contains the feature vectors of the elements that have paid their financial obligations on time and the second one called "Non-Solvency Region", which contains the feature vectors of the elements that have defaulted in paying their debts. Therefore, to predict creditworthiness borrowers, it is sufficient to identify which of the two regions includes his feature vectors; if it doesn’t correspond to any region, the credit decision-making requires further analysis. To develop and test our predictive proposed approach, a large set of real and recent credit data obtained from the UCI repository is used, we trained also on a real credit database from a Moroccan bank and the creditworthiness of borrowers are analyzed using two performance measurement indicators such as Classification accuracy and the AUC of the ROC curve as a robustness measurement criteria. The proposed model was compared to three traditional machine-learning algorithms: LR, RBF-NN and the MLP-NN. The experimental results show the improved performance of our proposed predictive method for predicting creditworthiness of borrowers.