提出算法FDMSP(fast distributed mining of sequential patterns),以解决分布式环境下的序列模式挖掘问题.首先对分布式环境下序列模式的性质进行了分析.算法采用前缀投影技术划分模式搜索空间,利用序列模式前缀指定选举站点统计序列的全局支持计数,利用局部约减、选举约减、计数约减等方法减少候选序列数,同时将算法分为3个子过程异步运行,使得算法具有较低的I/O开销、内存开销和通信开销,从而高效地生成全局序列模式.实验结果显示,在具有海量数据的局域网环境中,FDMSP算法的性能优于将数据集中后采用GSP算法68.5%～99.5%,并且FDMSP算法具有良好的可伸缩性.

Algorithm FDMSP (fast distributed mining of sequential patterns) is proposed in order to deal with mining sequential patterns in distributed environment and its properties are analyzed. The algorithm utilizes prefix-projected technique to divide the pattern searching space, utilizes polling site associated with prefix to get a global support, and utilizes local pruning, poll pruning and count pruning to decrease candidate sequences. It is divided into three sub-procedures which run asynchronously. As a result, the algorithm has lower I/O cost, memory cost and communication cost, and global sequential patterns are generated with higher efficiency. The experiments show that it outperforms the algorithm GSP after centralizing data by 68.5% to 99.5% and scaleable over LAN with huge amount of data.

Supported by the National Natural Science Foundation of China under Grant Nos.70171052, 90104030 (国家自然科学基金)