Research On The Building And Picking Of Stock Price Time Series ARCH Model

Since the foundation of Shanghai Stock Exchange,the Chinese stock market has developed over ten years, gradually from immaturity to maturity, it has became one of the most important capital markets in china. The past year, the Chinese stock market is up and down because of the impact of the international macroeconomic environment. For investors, it's more urgent to make the accurate analysis of the stock and make optimal investment decisions. For Chinese stock market regulators, it's also an arduous task to grasp the market dynamics, in order to make stock market develop healthily and steadily. So, whether managers or investors in the stock market have paid special attentions, especially, the analysis and prediction of the future stock market have became a hot research topic.Uncertainty research of the Stock market's price changes and empirical analysis are one of the core issues of the modern financial research. With the constantly improvement of Econometric Theory, in the actual economic activities, we often set up and use some econometric models and make a systematic and deep analysis on Chinese stock market. Meanwhile, econometric theory has promoted the development of the time-series methods, in 1982, Engle Professor proposed the famous Autoregressive Conditional Heteroscedasticity model (ARCH),hence, ARCH has been widely used in the stock market analysis and prediction research, which made good results. However, many current researches focused on sequence analysis, or simply choose a certain kind of model, the complete process to setting up ARCH model for time series until to firmed a fitting model for forecasting, especially ,while it existed a number of applicable models, how to choose the best model, ARCH model application is also worthy of study.Through analyzing the characteristic of ARCH model, and aimed at the situation of getting various fitting models which used existed methods, it suggested the basis of the model choosing which the fitted series is the closest the actual series, setting up integrated evaluation system including four statistical indexes which reflected statistical nature of series and five fitted indexes which reflected fitting effect, it suggested the method of measuring index weights due to independence among indexes based on index significance, and gave the TOPSIS for choosing the fitted model. We take Shanghai Automotive and Conch Profiles stock price time series profiles to make empirical analysis, use the above method to set up the most reasonable ARCH model, and make short-term predictions. By analyzing the model's short-term prediction results to proof the effectivity and rationality of the method proposed in this paper.