一种新颖的基于量化概念格的属性归纳算法

为了解决数据挖掘过程中挖掘的知识粒度过粗或过细问题，并利用概念格的偏序特性，提出了一种基于量化概念格的属性归纳算法．首先对概念格的外延进行量化，得到量化概念格，再根据概念格的哈斯图，采用概念的爬升进行相应的泛化，从而获得基于量化概念格的多层、多属性归纳．与面向属性归纳（AOI）算法相比较，结果表明所提算法不仅能实现AOI的单一属性归纳，还能进行多层、多属性的归纳，其属性泛化的路径不是惟一的，并且很容易在量化概念格的哈斯图中寻找合适的泛化路径和阈值，以此得到用户要求的、合理的属性归纳结果．

Novel Attribute Induction Algorithm Based on Quantized Concept Lattice

In order to deal with the problem of over-coarse or over-fine knowledge granularity in data mining,an attribute induction algorithm based on quantized concept lattice is proposed by using the partial property of the concept lattice.Firstly,the quantized concept lattice is defined by quantifying concept extension of the concept lattice,and then it is generalized using concept ascension according to the Hasse diagram of the concept lattice so as to get the induction with multi-level and multi-attribute based on the quantized concept lattice.Compared with the attribute-oriented induction(AOI) algorithm,the proposed algorithm can not only perform the(unitary) induction of AOI,but also carry out the induction with multi-level and multi-attribute,and the path of attribute generalization is not unique.Moreover,it is easy to find proper generalized paths and thresholds in Hasse diagram of quantized concept lattice to obtain the reasonable results required by users.