Title: Assessing model efficacy in forecasting EPS of Chinese firms using fundamental accounting variables: a comparative study
Authors: Qing Cao, Qiwei Gan, Marc J. Schniederjans
Addresses: Rawls College of Business, Texas Tech University, Lubbock, TX 79409-2101, USA. ' Rawls College of Business, Texas Tech University, Lubbock, TX 79409-2101, USA. ' Management, CBA 272, University of Nebraska-Lincoln, P.O. Box 880491, Lincoln, NE 68588-0491, USA
Abstract: In this paper, we compare the forecasting accuracy of two neural network models in forecasting earnings per share of Chinese listed companies based upon fundamental accounting variables. In one neural network model, weights estimated by back propagation were utilised, and in the other model a genetic algorithm was utilised. Based upon a sample of 723 Chinese companies in 22 industries over a ten year period, we found that the neural network model, using a genetic algorithm in forecasting, outperforms the neural network model with back propagation. Results also showed that the addition of fundamental accounting variables used in the neural network models further improved forecasting accuracy.
Keywords: neural networks; ANNs; neural network modelling; univariate model genetic algorithms; earnings per share; EPS; forecasting accuracy; Chinese financial markets; fundamental accounting variables; back propagation; comparative analysis; hypotheses testing; China; listed companies; modelling.
International Journal of Society Systems Science, 2010 Vol.2 No.3, pp.207 - 225
Available online: 02 Jun 2010 *Full-text access for editors Access for subscribers Purchase this article Comment on this article