Attribute Reduction Algorithm of Multi-Type Mixed Data Based on Multi-Kernel Fuzzy Conditional Entropy
Effective attribute reduction of data is a challenging task in data mining.At present,rough set theory is a common method to construct attribute reduction.However,the existing attribute reduction methods focus on single type data,not suitable for multi-type mixed data in real environment.In order to solve this problem,a multi-kernel fuzzy conditional entropy attribute reduction algorithm for multi-type mixed data is proposed.Firstly,a multi-kernel fuzzy similarity relation is proposed for the mixed multi-type data of nominal type,numerical type,interval type and set-valued type.Then,based on this multi-kernel fuzzy similarity relation,a multi-kernel fuzzy conditional entropy model is defined,and its monotonicity and boundedness are discussed.Finally,using the monotonicity of multi-kernel fuzzy conditional entropy,an attribute reduction algorithm for multi-type mixed data is proposed.The effectiveness of the algo-rithm is verified by the experimental analysis of UCI datasets.
国家科技期刊平台
NSTL
万方数据
