基于量化关联规则的敏感性分析  

Analysis of sensitivity based on quantitative association rules

作　　者：肖晗[1] 黄诚[1] 

机构地区：[1]西南石油大学计算机科学学院,成都610500

出　　处：《计算机应用》2017年第A01期255-257,291,共4页journal of Computer Applications

基　　金：国家自然科学基金资助项目(61379089)

摘　　要：通过数据挖掘产生的关联规则会存在大量无用和不感兴趣的规则,并且不能挖掘出元素之间的敏感程度,而敏感性分析方法效率低且合理性难验证,故提出一种使用量化关联规则进行敏感性分析的方法。该方法利用相对值概念描述数值型属性的变化程度及其对目标变量变化的影响,基于等宽分区完成相对值离散化,通过Apriori算法找出相互影响程度高的关联规则。最后通过分析某高校数据结构课程中的每个学生某个知识点的成绩对该方法进行了实证研究,得到了有价值的关联规则和选择知识块作为粒度分析挖掘出的结果最好。Data mining cannot dig out the sensitivity between elements, and may generate a lot of useless and uninteresting rules, and general sensitivity analysis method is inefficient and difficult to verify. To resolve the issue, a method of sensitivity analysis based on quantitative association rules was defined to find out association rules with high degree of mutual impact by Apriori algorithm. First, the relative value was used to describe the change degree of numerical attributes and the impact on the target variable. Second, the discretization of relative value was realized by equivalent width partitioning. At last the method was verified through the analysis of knowledge points score of data structure in a university, valuable association rules were obtained, and the conclusion was derived that it is best to choose the knowledge block as the analysis granularity.

关 键 词：数据挖掘 量化关联规则 敏感性 APRIORI算法 相对分值 数据结构 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]