摘要： 现有的时间序列异步周期模式挖掘方法是在获取1-pattern有效段及周期的基础上再以枚举法得到i-patterns，时间复杂度较高。为解决该问题，提出一种改进的异步周期模式挖掘方法。在时间序列符号化后，使用基于Sequitur的候选模式算法获取候选i-patterns及其事件位置序列，通过基于OEOP的i-patterns有效段生成算法得到1-pattern和i-patterns的有效段及周期，从而生成有效子序列。实验结果表明，该方法具有较高的挖掘效率。

关键词: 异步周期模式, Sequitur算法, 时间序列, 符号化, 数据挖掘

Abstract: Main existing algorithms mining asynchronous periodic patterns in time-series databases are designed to find i-patterns by enumeration based on the mining of valid segments and periodic patterns of 1-pattern, and time complexity is high. An improved algorithm based on Sequitur algorithm to mine asynchronous periodic pattern is proposed. It generates candidates for patterns based on the Sequitur algorithm to mine the recurring segments and their time-lists of events after the symbolization of time series. Then an algorithm generating valid segments of i-patterns based on One Event One Pattern(OEOP) algorithm is devised to mine directly some complex patterns, and finally valid subsequences are discovered. Experimental result demonstrates that this method has good mining efficiency.

Key words: asynchronous periodic pattern, Sequitur algorithm, time series, symbolization, data mining

中图分类号: 

• TP311