摘要
传统的特征选择算法需要依靠大量的数据进行有监督训练,具有高维和小样本的属性,进而造成数据冗余,导致维数灾难。针对这一情况,提出一种基于蚁群优化的特征基因选择算法。该算法对传统的蚁群算法模型进行优化,改进蚁群算法的参数选择方法,利用特征对不同数据集的敏感度,寻找最优基因,滤除无关基因,将特征选择过程转化为蚁群寻找最优路径的过程。实验表明,该算法可以有效地对特征进行优化选择,在降低数据维数的同时,提高分类的准确性和时效性。
The traditional feature selection algorithm relies on a large amount of data for supervised training,with high-dimensional and small-sample attributes,which in turn leads to dimensional redundancy caused by data redundancy.Aiming at this situation,a feature gene selection algorithm based on ant colony optimization is proposed.The algorithm optimizes the traditional ant colony algorithm model,improves the parameter selection method of ant colony algorithm,uses the sensitivity of features to different data sets,finds the optimal gene to filter out irrelevant genes,and transforms the feature selection process into the process of ant colony searching the optimal path.Experiments show that the algorithm can effectively optimize the feature selection and improve the accuracy and timeliness of the classification while reducing the dimension of the data set.
作者
侯远韶
HOU Yuanshao(Department of Mechanical and Electronic Engineering,Henan Industry and Trade Vocational College,Zhengzhou,Henan 451191,China)
出处
《中州大学学报》
2019年第6期120-123,共4页
Journal of Zhongzhou University
基金
河南省科技攻关计划项目“基于机器视觉的网状织物表面检测系统研制”(0721002210032)
河南省高等学校重点科研项目计划“基于多传感器信息融合的移动机器人最优路径规划策略研究”(20A120007)
关键词
蚁群算法
特征选择
优化
数据维数
Ant Colony Algorithm
feature selection
optimization
data dimension
