A Combination of Models for Financial Crisis Prediction: Integrating Probabilistic Neural Network with Back-Propagation based on Adaptive Boosting

DOI

10.2991/ijcis.2017.10.1.35How to use a DOI?

Keywords

Financial Crisis Prediction; Probabilistic Neural Network; Back-Propagation; Adaptive Boosting

Abstract

It is very important to enhance the accuracy of financial crisis prediction (FCP). Because the traditional probabilistic neural network (PNN) has some deficiencies, such as the difficult estimation of parameters and the high computational complexity, this paper proposes a new combination model, which combines back-propagation (BP) with PNN on the basis of adaptive boosting algorithm, to predict financial crisis. The BP algorithm is introduced to modify weights and smoothing parameters of PNN. In process of constructing BP-PNN models, the training set is divided into study and training samples to save the computational time. And the trained models are regarded as weak classifiers. Then these weak classifiers are integrated to constitute a stronger classifier by adaboost algorithm. To verify the superiority of the new model in terms of FCP, this article uses financial data of Chinese listed companies from Shenzhen and Shanghai Stock Exchange, and compares with adaboost BP, PNN and support vector machine models. The result shows that the new model has the highest prediction accuracy. Therefore, the new combination model is an excellent method to predict financial crisis.

Copyright

© 2017, the Authors. Published by Atlantis Press.

Open Access

This is an open access article under the CC BY-NC license (http://creativecommons.org/licences/by-nc/4.0/).

Download article (PDF)
View full text (HTML)

TY  - JOUR
AU  - Lu Wang
AU  - Chong Wu
PY  - 2017
DA  - 2017/01/01
TI  - A Combination of Models for Financial Crisis Prediction: Integrating Probabilistic Neural Network with Back-Propagation based on Adaptive Boosting
JO  - International Journal of Computational Intelligence Systems
SP  - 507
EP  - 520
VL  - 10
IS  - 1
SN  - 1875-6883
UR  - https://doi.org/10.2991/ijcis.2017.10.1.35
DO  - 10.2991/ijcis.2017.10.1.35
ID  - Wang2017
ER  -