基于邻近梯度的机器学习特征选择优化方法

Feature Selection Optimization Method of Machine Learning Based on Neighbor Gradient

针对当前机器学习特征选择方法存在运算时间较长、特征选择精准度较低的问题,提出邻近梯度的机器学习特征选择优化方法。根据机器学习概念,结合其特征选择原理特性,构建混合式特征算法;划分特征集并对其进行简易化,采用信息论和概率统计对特征子集分类,获得信息增益;根据特征间存在的特征关联性与类别属性之间的联系,对信息增益做离散化处理,完成特征选择;采用邻近梯度算法对特征选择进行优化。仿真结果表明,所提方法能够有效进行机器学习特征选择,并且通过优化可以提升特征选择的效率。

In current machine learning feature selection methods,the operation time is too long and the accuracy of feature selection is also low.Therefore,a method of optimizing the machine learning feature selection based on adjacent gradient was proposed.According to the concept of machine learning and the feature selection principle,a hybrid feature algorithm was designed.Moreover,the feature sets were classified and simplified,and then the feature subsets were classified by information theory and probability statistics,so that the information gain could be obtain.According to the relationship between the relevance of features and the class attributes,the information gain was discretized and feature selection was completed.Furthermore,the adjacent gradient algorithm was adopted to optimize the feature selection.Simulation results prove that the proposed method can effectively select the machine learning feature and improve the efficiency of feature selection through optimization.