 |
International Society of Science and Applied Technologies |
| Analysis of Loan Agreement using Deep Learning with a Regularization Term | ||||
| Author | Yutaka Azuma
|
|||
| Co-Author(s) | Masashi Kuwano; Yuka Minamino; Mio Hosoe
|
|||
| Abstract | Recent advances in information and communication technologies have resulted in the accumulation of vast amounts of data and the spread of machine learning methods. Deep learning (DL), which has attracted increasing attention among machine learning methods, is capable of highly accurate prediction; however, the black box nature of the computational process makes it difficult to interpret the prediction results. When using the results of DL-based analyses, an explanation of the rationale is required, and in recent years, transparency and accountability have become important requirements. In this study, we attempt to improve the interpretability of DL by selecting explanatory variables using a LassoNet model equipped with a variable extraction function. In an experiment, we predicted the probability of mortgage loan contracts using bank customer data comprising 344,384 samples and 141 variables. We also analyzed the relationship between the number of explanatory variables in the model and the prediction accuracy. By comparing the results of DL using only the important explanatory variables, we clarified the effect of selecting explanatory variables on the prediction accuracy.
|
|||
| Keywords | Deep Neural Network, LassoNet, Bank data | |||
| Article #: RQD2025-444 | ||||
Proceedings of 30th ISSAT International Conference on Reliability & Quality in Design |