A new time-series trend forecasting technique

In this paper, a new time-series trend forecasting technique based on rules and pattern is presented. For the given time series data, the technique is first to convert the time-series data into a time subseries data set, and then to excavate the time subseries data set and extract the underlying rules and patterns from them. And the rules are used as a tool to predict the trend of the time series. The mining algorithm is given last.