Research On TGARCH-M Stock Volatility Prediction Based On The Analysis Of Causal Hysteresis

With the economy developing rapidly, stock market is paid more attentions to by people. But it makes more difficulties for prediction that stock market’s high complexity and uncertainty. Many experts devote themselves to research on stock market’s prediction. And a better prediction model can not only describe the stock market’s volatility preferably, but also reduce the risk of investment.This dissertation mainly aimed at the low accuracy of model’s prediction caused by the inflection points’ hard prediction in stock price volatility. The analysis of inflection points in stock price volatility was based on the inconsistency phenomenon of price volatility and index volatility, and combined with the theory of Causal Bayesian Network with the TGARCH-M model for basal model. Researches were carried out in this article is as follows:Firstly, the stock price fluctuated acutely and the inflection point appeared continually for the reason that the stock price’s change was complex and uncertain. For this reason, a kind of generalized autoregressive conditional heteroscedastic model based on characteristics of hysteresis (LRD-TGARCH-M) was proposed for inflection points’ prediction. Hysteresis was defined based on the inconsistency phenomenon of price volatility and index volatility, and the hysteresis degree calculation model was proposed through the energy volatility of the stock. The hysteresis cumulative was put into the average share price equation and the variance equation of TGARCH-M model to help forecast the inflection points and improve the accuracy of prediction.Secondly, the further analysis based on the basis of LRD-TGARCH-M model showed that the characteristics of hysteresis for calculating the hysteresis cumulative degree had influence with each other. For this reason, a kind of generalized autoregressive conditional heteroscedastic model based on causal hysteresis (CLD-TGARCH-M) was proposed. Firstly, a Bayesian network was built using the characteristics of hysteresis, and then the local causal structure of this Bayesian network was studied by disturbance. The hysteresis degree was amended by the local causal structure of this Bayesian network. Finally, the correctional hysteresis degree was put into the TGARCH-M model for improving accuracy of prediction.Finally, the contrastive analysis on the two algorithms respectively was carried out through the Shanghai composite index data. The results show that the two algorithms both can improve the accuracy of prediction, and the second proposed algorithm has greater advantage than the first proposed algorithm for inflection points’prediction.