Συγκριτική μελέτη μεθόδων μηχανικής μάθησης για πρόβλεψη πιστωτικού κινδύνου
A comparative study of machine learning approaches for credit risk prediction
View/
Keywords
Credit risk prediction ; Random forests ; Support vector machines ; Μηχανική μάθηση ; Νευρωνικά δίκτυα ; Πιστωτικός κίνδυνοςAbstract
Credit risk forecasting is an important area in finance as it allows financial companies to assess the risk associated with extending credit to customers. Credit risk analysis includes the measure to investigate the applicant's likelihood of repaying the loan on time and predicting his inability to repay. There are two main risks, the loss of potential revenue resulting from not approving one good candidate or rejecting many, and the financial loss resulting from approving a candidate who ends up defaulting on the loan.
Machine learning methods have been developed and successfully applied in this field to predict the likelihood of borrowers becoming credit distressed. In our work, we will use three methods, Artificial Neural Networks, Support Vector Machines (SVM) and Random Forests. With the help of these algorithms we will train a dataset that includes information about borrowers and whether they have responded to their loan payments. This way of predicting credit risk can provide important information on loan performance and borrowers' ability to pay.