Deep learning for repayment prediction in leasing companies

Abstract

Purpose: This paper aims to improve repayment prediction in leasing companies using a deep learning model. Design/Methodology/Approach: In this work, we prepare some deep learning models and compare them with other solutions based on artificial intelligence like, multiple regression, decision tree, random forest, and bagging classifier. Findings: The developed model enables automatic analysis of large amounts of data that changes quickly and is often unstructured. Additionally, the input vectors consist of specific attributes related to leasing. The results of experiments allow us to conclude that the prediction accuracy of the developed model is higher than reference models used currently in leasing companies. Practical Implications: The developed model has recently been implemented in the Decision Engine system (a system used by leasing companies in Poland) developed by BI Technologies Sp. Z o.o. Company. Originality/Value: Financial institutions automate and simplify credit procedures, eliminating the analyst from the process and replacing him with automatic decision-making processes based on a scoring or similar models. However, to automatically analyze the significance of phenomena occurring in the environment of organizations that affect the assessment of customer's repayments, it is necessary to use artificial intelligence tools.