摘要
逻辑回归是机器学习中一个非常常用且经典的分类模型。提出了一种逻辑回归的敏感性计算方式,并将其应用在特征选择领域中。以机器学习中的iris数据集为例,首先给出逻辑回归的输出对输入变量的敏感性定义,然后选择敏感性值比较高的属性特征组成特征子集,最后利用分类器对选取的特征进行验证。通过实验结果表明,利用敏感性方法能够选择出有效的特征,该方法是有效和可行的。
Logistic regression is a very popular and classic classification model in machine learning. In this paper, a sensitivity calculation method of logistic regression is proposed and applied in the field of feature selection. Using the iris data set in machine learning as an example, Firstly, the sensitivity of the output of logistic regression to the input variables is given. Secondly, the attribute features with relatively high sensitivity values are selected to form a feature subset. Finally the selected features are verified by the classifier. The experimental results show that the effective features can be selected by using the sensitivity method, and the method is effective and feasible.
作者
王凌妍
张鑫雨
许胜楠
王禹力
甄志龙
WANG Lingyan;ZHANG Xinyu;XU Shengnan;WANG Yuli;ZHEN Zhilong(Tonghua Normal University,Tonghua,Jilin 134002,China)
出处
《信息记录材料》
2022年第7期30-33,共4页
Information Recording Materials
基金
通化师范学院省级大学生创新创业训练计划项目(S202110202063)
吉林省教育厅科学研究项目(JJKH20210533KJ)。
关键词
逻辑回归
敏感性
特征选择
Logistic regression
Sensitivity
Feature selection
作者简介
王凌妍(1999-),女,河南南阳,本科,研究方向:计算机应用;通信作者:甄志龙。
