最大频繁序列挖掘是数据挖掘的重要内容之一。在深入分析频繁序列特点以及已有序列挖掘算法的基础上,提出一种新的最大序列挖掘算法Huffman-MaXseq.与传统的“候选最大频繁序列集生成——测试”思路不同,该算法采用“边生成候选序列边测试”的思想,从而有效地减少了候选序列的生成。该算法基于构造哈夫曼树(最优树)的方法,对每个序列赋予权值,按权值的大小选取序列,连接生成新的候选频繁序列,再产生最大频繁序列。
Mining maximal frequent sequences is an important topic in the data mining research. On the basis of analyzing the characters of frequent sequences and the algorithms for mining frequent sequences developed previously, this paper propose a new algorithm for mining maximal frequent sequences, namely Huffman-MaxSeq. Different from traditional method of “generating candidate maximal frequent sequences, then testing for qualification”, the algorithm use the idea of “testing for qualification while generating candidate sequences”, effectively reducing the number of redundant frequent candidates generated. Based on Huffman-tree structure, the algorithm assigns a weight to each sequence, selects sequences according to their weight, generates the new candidate frequent sequences by joining these sequences, and finally generates the maximal frequent sequences.