Mining frequent sub-trends in time-series databases

Mining time-series databases is a novel and important problem in the field of data mining. Most previous work focused on the similarity of the naive time series. Some authors proposed the similarity of trends rather than time series. Based on the approach presented in this paper, more formal definitions of trends for process data are given. The problem of mining frequent sub-trends in a long trend sequence is formulated and an algorithm to solve this problem is developed. Experiments were done on a simplified simulation system, which showed that the satisfying results were achieved.