Abstract
Risk control has always been a major challenge in finance. Overdue repayment is a frequently encountered discreditable behavior in online lending. Motivated by the powerful capabilities of deep neural networks, we propose a fusion deep learning approach, namely AD-MBLSTM, based on the deep neural network (DNN), multi-layer bi-directional long short-term memory (LSTM) (BiLSTM) and the attention mechanism for overdue repayment behavior forecasting according to historical repayment records. Furthermore, we present a novel feature derivation and selection method for the procedure of data preprocessing. Visualization and interpretability improvement work is also implemented to explore the critical time points and causes of overdue repayment behavior. In addition, we present a new dataset originating from a practical application scenario in online lending. We evaluate our proposed framework on the dataset and compare the performance with various general machine learning models and neural network models. Comparison results and the ablation study demonstrate that our proposed model outperforms many effective general machine learning models by a large margin, and each indispensable sub-component takes an active role.
Highlights
With the development of the economy and the rising level of consumption in national standards of living, the majority of people and companies have encountered capital turnover problems and have attempted to obtain a loan for consumption, capital turnover, investment, etc
We propose an overdue repayment forecasting method based on a fusion of deep learning models
We introduce deep learning models into the domain of risk control in online lending; overdue repayment forecasting based on historical repayment behaviors
Summary
With the development of the economy and the rising level of consumption in national standards of living, the majority of people and companies have encountered capital turnover problems and have attempted to obtain a loan for consumption, capital turnover, investment, etc. Sci. 2020, 10, 8491 handling small monetary loans with high frequency is difficult in online lending Once frauds such as overdue repayment occur, tracking accountability and recuperating loss become difficult problems. Especially sequential neural networks, have achieved remarkable performance in event sequence-related tasks [3,4,5]. These kinds of models are well suited to handle the massive amount of online repayment behavior data. We introduce deep learning models into the domain of risk control in online lending; overdue repayment forecasting based on historical repayment behaviors. We visualize differentiated attention weights to explore the key event time steps and analyze the feature importance of derived features to determine the causes of overdue repayment
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
