多支持度下用户行为序列模式挖掘方法研究

RESEARCH ON MINING USER BEHAVIOR SEQUENTIAL PATTERNS WITH MULTIPLE SUPPORTS

针对现有用户行为序列模式挖掘方法的单一支持度局限性问题,提出一种基于前缀树结构的多支持度序列模式挖掘方法。设计一种多支持度条件下的前缀树结构MSLP-tree,并基于此结构提出一种序列模式增长算法MSLP-growth。通过考虑各数据项不同最小支持度,获取更精确的频繁序列模式,在确保挖掘结果的准确性和完整性的前提下,大大压缩搜索空间,缩短挖掘时间。实验结果表明,相较于MS-GSP算法,MSLP-growth算法具有更高的挖掘效率和可扩展性。

Aiming at the limitation of the single support threshold of most current sequential pattern mining methods, a sequential pattern mining method with multiple minimum supports based on prefix tree structure was proposed. We designed a prefix tree structure MSLP-tree with multiple supports and proposed a sequential pattern growth algorithm based on this structure, which was called MSLP-growth. The algorithm obtained the precise frequent sequential patterns by considering the varied minimum support of each item. In the premise of ensuring the accuracy and integrity of the mining results, the storage space was greatly reduced and the search time was shortened. Experimental results showed that compared with MS-GSP algorithm, MSLP-growth algorithm had higher mining efficiency and scalability.