摘要
针对数据缺失这一普遍情况,提出一种改进的微粒群优化特征选择方法.首先,采用多重插补方法对缺失的数据进行插补,得到完整数据集;然后,采用k折交叉验证法计算分类器的精度,并在算法运行后期,对微粒群进行K均值聚类,从中选择微粒的全局最优点;最后,通过UCI中4个典型测试问题,仿真验证了所提算法的有效性.
In view of the general situation of missing data, an improved feature selection method for particle swarm optimization is proposed in this paper. Firstly, multiple interpolations are used to interpolate the missing data and get a complete set of data. Then, the K fold cross validation method is used to calculate the accuracy of the classifier. After the operation of the algorithm, the K-means cluster is used to cluster the particle swarm to select the best global advantage of the particle. Finally, the effectiveness of the proposed algorithm is verified by 4 typical test problems in user computer interface （UCI）.
作者
胡滢
HU Ying(Department of Electrical Engineering, Tongling University, Tongling 244061, Chin)
出处
《西安文理学院学报（自然科学版）》
2018年第2期80-83,共4页
Journal of Xi’an University(Natural Science Edition)
基金
铜陵学院校级科研项目:数据缺失下微粒群优化特征选择方法(2016tlxy24)
关键词
微粒群优化
特征选择
多重插补
K均值聚类
particle swarm optimization
feature selection
multiple imputation
K-means clustering
