Comparative Study on Risk Early Warning Models of Capital Chain in Real Estate Industry ‐‐ Based on BP Neural Network Model and SVM

Jingxuan Liu ； Yijun Chen ； Yu Gan

《International Journal of Social Science and Education Research》 5卷10期 (2022/10) Pp. 71-78

https://doi.org/10.6918/IJOSSER.202210_5(10).0012

摘要

This paper takes the financial status of the capital chain of listed companies in the A-share real estate industry from 2016 to 2021 as the research object, constructs an early warning indicator system from four dimensions of capital chain investment chain, operation chain, return chain and non‐financial indicators, using BP neural network , Support Vector Machine (SVM) to build early warning models respectively, and divide the early warning time into two groups of variables: 1‐3 years short‐term early warning and 3‐5 years long‐term early warning, and then choose a model that is more suitable for the real estate industry. . The experimental results show that the difference in prediction accuracy between the two models is small in short‐term early warning, but in long‐term early warning, the prediction accuracy of BP neural network model is significantly better than that of SVM.

關鍵字

BP Neural Network ； Support Vector Machine ； Capital Chain Risk Early Warning

參考文獻

Jingxuan Liu, Yijun Chen. An Empirical Study of Financial Early Warning in the Real Estate Industry-- Based on the Principal Component Logistic Regression Model[J]. World Scientific Research Journal,2022,8(1).

Altman E.I.Financial Ratios Discriminate Analysis and Prediction of Corporate Bankruptcy[J]. The Journal of Finance,1968,(4):589-609

Ohlson,J.S.Financial Ratios and the Probabilistic Prediction of Bankruptcy [J].Journal of Accounting Research,1980,(19):109-131.

Adnan, Aziz, Humayon. Predicting corporate bankruptcy: where we stand?[J].Corporate Governance The international journal ofbusiness in society,2006,6(1):18-33.

Lin T H. A cross model study of corporate financial distress prediction in Taiwan: Multiple di scriminant analysis, logit, probit and neural networks models[J]. Neurocomputing, 2009, 72(16) 3507-351.