一种基于模糊聚类和蕴涵度的模糊关联规则挖掘算法

An Algorithm of Fuzzy Association Rules Mining Based on Fuzzy Clustering and Implication Degree

对于模糊关联规则挖掘算法存在的不足,首先为了软化数量型属性论域的划分边界,借用了FCM算法将数量型属性离散化,并把数据集划分成若干个模糊集等级;然后,对模糊置信度进行定义时,把经典关联规则中的置信度的定义经过扩展后直接运用到模糊集上,不免会带来一些逻辑推理上的问题,采取了蕴涵度代替模糊置信度的方法,引入模糊蕴涵算子,经过进一步推理论证,证明了蕴涵度能够用模糊支持度来代替。提出了一种基于模糊聚类和蕴涵度的模糊关联规则挖掘算法,并通过实验证明了算法的有效性。

Aiming at the deficiency of the algorithm of fuzzy association rules mining, in order to soften the domain partition boundary of the quantitative attributes, the quantitative attributes are partitioned into several fuzzy sets by fuzzy C-means algorithm; then, the extended definition of confidence degree in classical association rule is directly used in fuzzy sets when defining fuzzy confidence degree, some problems of logical reasoning may occur, so the method of implication degree is used instead of fuzzy confidence degree, and fuzzy implication operator is introduced. Through further inference and proof, it is shown that implication degree can be replaced with fuzzy support degree. An algorithm of fuzzy association rules mining based on fuzzy clustering and implication degree is proposed, and an experiment shows that the algorithm is effective.