Stock market is full of risk. The author attempts to forecast stock price and guide the investment effectively by finding the potential rules of securities investment risk, which is a timely waved series. The thesis puts forward a new securities investment risk definition based on the average, the variance, and the frequency of profit and loss of the yield in order to educe an index which is used to measure a risk yield of single bond investment.After discussing the non-linear characteristic of the investment risk series, the thesis investigates several forecasting methods for non-linear risk series including AR(P) model, ARCH model and neural network model.In this paper, the author makes use of the historic data of Shang Hai Securities Exchange, which are calculated by new risk measure index and forms a risk yield series which is an non-linear series to carry through the demonstration research by five forecasting methods. The author summarizes a lot of useful laws of the model-building after analyzing the forecasting results to verify that combination forecasting is superior in improving forecasting precision and veracity.At the end of this paper, the author generalizes the contents of this thesis, pointing out the development direction of this field.
|