基于树结构多重最小支持度的挖掘算法研究

摘要

传统的序列数据库中各数据项的最小支持度是单一的,且不能有效挖掘用户感兴趣的、稀有的数据项。为了有效提高数据挖掘的效率和准确率,文中基于PLWAP-tree提出了前序链接多重支持度树( Preorder Linked Multiple Supports tree,PLMS-tree)来存储序列数据库,并进一步提出了多重最小支持度条件模式增长( Multiple Support-Conditional Pattern growth,MSCP-growth)算法。算法采用对每个频繁数据项设置多重最小支持度的方法来减少空间和时间的开销。对于每个频繁数据项设置不同的支持度,来挖掘用户所需的数据序列,能有效提高数据挖掘的效率和准确率。实验结果验证了算法的有效性,对序列模式下的数据项目集挖掘的时间效率和空间效率有明显的提高。%The minimum support of each data in traditional sequence database is single and cannot effectively mine user interest and rare items. In order to effectively improve the efficiency and accuracy of data mining,based on PLWAP-tree,the Preorder Link Multiple Sup-ports tree ( PLMS-tree) is structured to store the entire sequence databases,and then the algorithm of Multiple Supports Conditional Pat-tern growth ( MSCP-growth) is proposed. It reduces the cost of space and time through multiple minimum supports. Each frequent item sets different supports to meet the required data pattern. Experimental results demonstrate the effectiveness of the algorithm,which can sig-nificantly improve mining time and space efficiency for the data item set in the sequence mode.