摘要
传统的基于关联规则的挖掘算法采用的是统一的最小支持度,但是在实际的事务数据库中数据项的重要性是不同的。针对目前多支持度和增量式关联规则更新维护的局限性,提出一种基于多支持度的增量式关联规则挖掘算法。允许用户根据不同项的重要性设置权值,有利于发现更多有趣的规则。采用矩阵的向量内积策略,结合动态剪枝,无需多次扫描事务数据库,不生成庞大候选集。实验结果验证了算法的有效性。
The uniform minimum support degree is proverbially adopted in traditional mining algorithms based on association rules. However, the significance of various data items is different in the transaction database in reality. In view of the limitations of multiple minimum supports and incremental association rules mining, we put forward the incremental updating algorithm of association rules mining based on multiple supports. The approach, which allowed users to set different weights according to the significance of items, would make it easier to find more interesting rules. Moreover, the new proposed method dispensed with a complex multiple scanning process due to the strategy of vector inner product and dynamic pruning. Consequently, there might not be enormous candidate sets. Experiment was carried out to illustrate the validity of the method.
出处
《南昌大学学报（理科版）》
CAS
北大核心
2015年第2期139-142,共4页
Journal of Nanchang University(Natural Science)
基金
国家自然科学基金资助项目(61070139)
关键词
多支持度
关联规则
增量更新
数据挖掘
multiple supports
association rules
incremental updating
data mining
