The tradional forecasting methods on time seires are either based on the whole data set or by adding a damped vibration to reflect the temporal importance of data. In this paper, we come up with a new forecasting method based on sliding window. Our method sets up a proper ate window for keeping the latest and newest data, and use the latest N tuples to forecast. When data come into the window, the algorithm can compute the regression parameters by utilizing the results which were produced before. The experimental results show that our algorithm has better effectiveness and forecasting veracity than the traditional one.
展开▼