摘要: 相关度是关联规则挖掘中衡量一条规则是否有趣的重要的参数。该文通过对传统的相关度计算公式的分析,指出了传统计算方法存在不对称性和不完备性两个缺陷,提出了提升率和影响率两个概念。通过理论分析和实验验证,证明了采用提升率和影响率两个参数来描述相关度与传统的描述方法相比,结果更加准确可靠。
关键词:
关联规则,
相关度,
作用度,
提升率,
影响率
Abstract: Relative degree is an important parameter in evaluating a rule whether it is interesting or not in association rules mining. Different from previous research, the paper focuses on the importance of upgrade rate and influence ratio in the study of relative degree calculation and points out the dissymmetry and imperfection of the traditional calculation method. Theoretical analysis and experiments show that applying upgrade rate and influence ratio to relative degree calculation is more accurate and reliable than traditional method.
Key words:
Association rules,
Relative degree,
Lift,
Upgrade rate,
Influence ratio
中图分类号:
马占欣;黄维通;陆玉昌. 相关度计算方法存在的问题及修正[J]. 计算机工程, 2007, 33(11): 67-69.
MA Zhanxin; HUANG Weitong; LU Yuchang. Problems on Calculation Method of Relative Degree and Its Modification[J]. Computer Engineering, 2007, 33(11): 67-69.