基于自适应邻域空间粗糙集模型的直觉模糊熵特征选择

特征选择是数据预处理中一项很重要的技术，主要从原始数据集的特征中选出一些最有效的特征以降低数据集的维度，从而提高学习算法性能.目前基于邻域粗糙集模型的特征选择算法中，由于没有考虑数据分布不均的问题，对象的邻域存在一定的缺陷.为了解决这个问题，采用方差来度量数据的分布情况，重新定义二元邻域空间，基于此提出自适应二元邻域空间的粗糙集模型，并将该模型与邻域直觉模糊熵结合作为特征评估的方式，进而构造相应的特征选择算法.UCI实验结果表明：所提出的算法能够选出更小且具有更高分类精度的特征子集，同时算法拥有更少的时间消耗.因此所提的特征选择算法具有更强的优越性.

Intuitionistic Fuzzy Entropy Feature Selection Algorithm Based on Adaptive Neighborhood Space Rough Set Model

Feature selection is a very vital technology in data preprocessing.In this method, some most effective features are mainly selected from the features of original data sets, which is aimed to reduce the dimension of data sets.Accordingly, the performance of the learning algorithm can be improved.In the feature selection algorithms based on the neighborhood rough set model, without considering data uneven distribution, there currently exist some defects in the neighborhood of object.To solve this problem of data uneven distribution, variance is adapted to measure the distribution of the data, and the binary neighborhood space is redefined, then the rough set model of the adaptive binary neighborhood space is proposed according to this binary neighborhood space.As well as, the new rough set model of the adaptive binary neighborhood space is combined with the neighborhood intuitionistic fuzzy entropy as the method of the evaluation of features, and then the corresponding feature selection algorithm is also constructed.The experimental results of UCI show that the proposed intuitionistic fuzzy entropy feature selection algorithm can select smaller feature subsets which have higher accuracy of classification, at the same time, the intuitionistic fuzzy entropy feature selection algorithm based on adaptive neighborhood space rough set model also has less time consumption.Therefore, the proposed feature selection algorithm has stronger superiority in this paper.