基于矩阵算法的序列模式挖掘研究  

Research on Sequential Pattern Mining Based on Algorithm of Matrix

作　　者：卢博礼[1] 张小平[1] 王翰虎[1] 

机构地区：[1]贵州大学计算机科学与信息学院,贵州贵阳550025

出　　处：《计算机技术与发展》2011年第1期120-124,共5页Computer Technology and Development

基　　金：贵州省2008年省级信息化专项资金项目(0830);贵州省科技计划工业攻关课题(黔科合GY字[2008]3035)

摘　　要：序列模式挖掘中几种算法的缺点:都要进行多次扫描数据库,CPU要进行多次I/O操作。这成为序列挖掘中的一大瓶颈,使得算法在实际应用中的效率不高。文中提出一种矩阵算法,即在一次扫描数据库时,根据扫描数据建立由0和1组成的事务矩阵。接下来的大序列、序列模式等都是通过矩阵的列向量对应元素的相乘运算和简单的加法运算而得到。从而使算法得到进一步优化,提高了CPU的使用率,解决了序列挖掘中的瓶颈问题。本算法通过大量的数据实验,证明了算法确实有效地优化了算法的时间复杂度。Based on several shortcomings of algorithm of the sequential pattern mining. Itmust scan the database many times, so a number of CPU to I / O operation, it become bottlenecks. The efficiency of algorithm is not high in practical applications. The paper presents an algorithm of matrix, that is, in a scan database, in accordance with scan data to establish the matrix of affairs which are composition of 0 and 1. Then, large sequence and sequential patterns are all out pass the vector of matrix multiplication operator corresponding to the elements and simple addition operations have been. So that the algorithm has been further optimized to improve the CPU rate of utilization. The algorithm use a large number of experimental data to prove that the algorithm is indeed effective to optimize the algorithm of complexity of time.

关 键 词：序列模式挖掘 序列模式 大序列 矩阵算法 连接运算 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]