事件序列上的频繁情节挖掘算法

摘要

Mining frequent episodes on the event sequences is one of the hot areas of data mining. In this paper, support based on non-overlapped occurrence is definited. We propose an algorithm called NONEPI++ for mining frequent episodes, which firstly generate candidate episodes by join episodes, then generate frequent episodes by pre-pruning and timestamp calculating. The algorithm can improve the efficiency of mining episodes. Experiments show that NONEPI++ algorithm can effectively mine frequent episodes.%事件序列上的频繁情节挖掘是时序数据挖掘领域的热点之一，基于非重叠发生的支持度定义，提出一个频繁情节挖掘算法NONEPI++，该算法首先通过情节串接产生候选情节，然后通过预剪枝和计算情节发生的时间戳来产生频繁情节。算法只需扫描事件序列一次，大大提高了情节挖掘的效率。实验证明， NONEPI++算法能有效地挖掘频繁情节。