Welcome to Francis Academic Press

Academic Journal of Computing & Information Science, 2022, 5(6); doi: 10.25236/AJCIS.2022.050615.

A study of bank failure risk based on neural network model

Author(s)

Mengyi Kang1, Kang Shi2, Menghan Ma3

Corresponding Author:
Mengyi Kang
Affiliation(s)

1School of Mathematics and Statistics, Anhui Normal University, Wuhu 241003, Anhui, China

2School of Information Science and Technology, Shijiazhuang Tiedao University, Shijiazhuang 050043, Hebei, China

3School of Architecture and Construction, Shijiazhuang Tiedao University, Shijiazhuang 050043, Hebei, China

Abstract

This paper focuses on the analysis of bank failure causes i.e. prediction. Based on the neural network model, we use principal component analysis and fuzzy C-mean cluster analysis algorithm to achieve the evaluation of bank efficiency, the mining of bank insolvency causes, and the prediction of bank failure risk. By considering the literature search and production method together, we classify the 64 indicators of the indicator set into input indicator layer and output indicator layer. In the process of data analysis, we found that some of the higher weights do not objectively reflect the causes of bank failures. In order to avoid simply selecting 5 indicators from the 64 indicators, we used principal component analysis to reduce the dimensionality, set the number of extracted factors to 5, and consider the bank failure factors comprehensively through the 5 selected principal components, and calculate the percentage weights of the eigenvalues calculated for each principal component each year to complete the calculation of each bank's score. Whether a bank fails or not has a strong nonlinear relationship with 64 indicators, and considering that some of the data present exponential relationship, logarithmic relationship, less pure linear relationship between the data, and the indicator dimension is too large. Therefore, we decided to build a model based on BP neural network model, and continuously reduce the error by forward, backward transmission.

Keywords

Entropy weight method, principal component analysis, BP neural network, fuzzy C-value clustering analysis

Cite This Paper

Mengyi Kang, Kang Shi, Menghan Ma. A study of bank failure risk based on neural network model. Academic Journal of Computing & Information Science (2022), Vol. 5, Issue 6: 101-107. https://doi.org/10.25236/AJCIS.2022.050615.

References

[1] Li H.M., Qin G.H.,Hao B., Ji J.Z.. Matching degree calculation method in fuzzy matching[J]. Computer Engineering, 2010,36(06):184-185.

[2] Tian Xiaona,Dong Jing. Fuzzy C-mean clustering for remote sensing image classification[J]. Mine Survey, 2011(03):32-34+36.

[3] Wen Chuanjun,Zhan Yongzhao,Ke Jia. Generalized equilibrium fuzzy C-mean clustering algorithm[J]. Systems Engineering Theory and Practice, 2012,32(12):2751-2755.

[4] Ji G. Stock price trend prediction based on BP neural network[D]. Zhengzhou University, 2018.

[5] Zhang Peng. Comprehensive evaluation research based on principal component analysis [D]. Nanjing University of Science and Technology, 2004.