Search

Stock market forecasting has attracted a lot of research interests in previous literature. Traditionally, the autoregressive moving average (ARMA) model has been one of the most widely used linear models in time series forecasting. However, the ARMA model cannot easily capture the nonlinear patterns. And recent studies have shown that artificial neural networks (ANN) method achieved better performance than traditional statistical ones. ANN approaches have, however, suffered from difficulties with generalization, producing models that can overfit the data. Support vector machines (SVMs), a novel neural network technique, have been successfully applied in solving nonlinear regression estimation problems. Therefore, this investigation proposes a hybrid methodology that exploits the unique strength of the ARMA model and the SVMs model in the stock market forecasting problem in an attempt to provide a model with better explanatory power. Real data sets of stock market were used to examine the forecasting accuracy of the proposed model. The results of computational tests are very promising.

The paper separates the investors in stock market into informed traders and uninformed traders according to the characteristic of the investor heterogeneity, which denies the classical capital market theories' assumption that investors are homogenous. Firstly, the assumption is made that two kinds of investors have respective forms of behavioral decision models which the irrational expectation and the irrational risk aversion are leaded into. Secondly, the difference of reaction strength and behavioral pattern is tested between the two kinds of investors by Event Study Methodology. Finally, verify the rationality of binary classification in behavioral decision model. It has been found through empirical research on Chinapsilas capital market that a body of evidence with systematic patterns can not be easily explained with rational asset pricing models.