摘要
提出了一种新的数据库划分方法。该方法应用于需要产生候选项的频繁模式的挖掘过程,可以大大减少对数据库的扫描操作,提高数据挖掘效率,特别是对于较长模式的数据挖掘更是如此。该方法是将交易数据库按照交易的长度(或者说模式的长度)划分成若干个子数据库,将等长度的交易划分到同一个子数据库中,这样在获取候选项的支持度时,只需要扫描模式长度大于等于相应候选项长度的子数据库即可,从而减少了对数据库的扫描操作。给出了基于数据库划分的挖掘算法,通过理论推导和实验证明了该方法的有效性。
A new database partition method for mining frequent patterns is proposed. In the process of mining frequent patterns, based on this method, the number of records to be scanned can be greatly reduced, especially for mining long patterns. In this method, database is partitioned according to the length of transactions or patterns. Transactions with the same length are partitioned to the same subdatabase. Therefore, it only need scan those subdatabases in which the transactions are longer than or equal to the candidate item set when calculating the support number of a candidate item set. As a result, the number of scanned database records will be reduced. The mining algorithm based on database partition is put forward, and the validity of the algorithm is proved by theoretical deduction and experiment.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2004年第11期1666-1668,1745,共4页
Systems Engineering and Electronics
基金
江苏省高校自然科学研究计划(03KJD110089)
航空基金(01F52036)资助课题
关键词
数据挖掘
频繁模式
数据库划分
data mining
frequent pattern
database partition