The Study Of Short-term Trend Information Model In Time Series

Time series exist widely in either nature or society. Data is in the form of time series in many domains such as economic, finance, engineer, technology, astronomy, geography, meteorology or biology. People can know things more accurate. So it is essential for making correct decision.From late 20th century on, the study on time series forecasting made steadily progress. Many methods depend on financial theory and then construct forecasting model. Others use different statistical methods and time series models. With the development of computer science and more data, we can use more models, and the efficiency of models is increasing. With the development of financial market in our country, financial time series is becoming more important for financial investors. More and more scholars and investors pay attention to the study of time series. In financial market investors pay more attention to the trend of time series than the accurate value of time series. Because the investors if only know the trend of time series, they will benefit from financial market. This thesis is on the trend of financial time series. We detect the short term signals and predict the short term trend. Then based on the short term trend we predict the long term trend of the financial time series. The work of this thesis is as follows:1,Put forward a method to estimate short term trend. Short term trend is the rising or falling trend in time series. In this paper we definite short term trend as this if the time series fall or rise continuously and then rise or fall three times, that is rising trend and falling trend respectively.2,Put forward a short term trend model based on signal. We use ordinary least square to build model for short term trend.3,Put forward a method to estimate long term trend. We decompound the time series to trend and disturbance term. The trend model has been built through short term signal. The model of disturbance is built using ARMA model. The long term trend model is the summation of two factors.4. We test the short term model. The test of Shanghai A stock market shows the model is effective. The result of media-long term trend model is able to reduce the error of short term model. So the model can guide the investors in financial market.