基于数据流的频繁集挖掘

Discovering Frequent Itemsets over Data Streams

针对数据流特殊的数据类型,提出了一种新的数据流挖掘算法.该算法引入了一个全新的优化方法,将边界集和频繁产生集结合起来.频繁产生集是频繁集的一种无损简缩表达方式.它所包含的模式数量比频繁集所包含的模式数量小若干数量级.边界集是频繁产生模式和其他模式之间的边界,通过观察边界集的变化可以生成新的频繁产生模式.实验结果表明,该算法的性能有明显的提高.

This paper presented a novel algorithm to discover frequent itemsets over data streams. The algorithm introduces a novel optimization technique combining with border sets and generator representation. The generator representation is a kind of lossless and concise representation of the set of frequent itemsets. It has smaller orders of magnitude than the set of all frequent itemsets. Border sets are the borderline between the frequent generators and other itemsets. New generators can be found through monitoring border sets. The experimental results show the improved performance when compared with the exist- ing algorithms over data streams.