多数据流时间序列中的依赖模式发现算法研究

Research on The Structure Patterns of the Multiple Time Series

针对多数据流组成的时间序列中发现有用的结构模式的MSDD算法不能很好地对节点剪枝,以及直观地表示模式的时间关系的问题,经过研究,提出了发现多数据流时间序列结构模式的算法:时间窗口移动筛选算法(TWMA).采用事件序列化的策略来发现多流时间序列中的依赖模式,与MSDD相比,在表示上更直观,发现模式的过程更灵活.

It is very important to find the structure patterns from the multiple time series. A famous algorithm provided by Tim oates is the MSDD, it finds the dependency patterns by the dependency trees. The main problem is that it cant have a trim in times to the nodes, and it can't express the time relation clearly. A TWMA(Time Window Move Algorithm) algorithm is made to solve the above problems. It is more simple, clear, concision than the other algorithm by use of the fuzzy theory and the ideas of events serial. 