摘要
针对常见属性约简算法常会误删除具有实际意义属性这一问题,文章提出一种改进的属性约简算法。该算法通过对权重低属性进行隐含关系挖掘,找出具有隐含关系的属性进行分析融合,得到新的属性。属性约简算法在保证对属性约简的同时还能保留属性的实际意义。实验结果表明,新的属性约简算法在UCI公共数据集上得到的分类结果相较于传统随机森林算法和SVM算法精度分别提升4%和7%;在驾驶行为数据集上得到的结果相较于传统随机森林算法和SVM算法分类精度分别提升6%和8%。
In order to solve the problem that common attribute reduction algorithms usually delete the attributes with practical significance,this paper proposes an improved attribute reduction algorithm.The algorithm mines the implicit relationship of the attribute with low weight,finds the attribute with the implicit relationship and analyzes and fuses it,and obtains the new attribute.While ensuring the reduction of attributes,the actual meaning of the attributes can be preserved.Experimental results show that the new attribute reduction algorithm improves the classification results on UCI public datasets by 4%and 7%,respectively,compared with the traditional random forest and SVM algorithms.Compared with the traditional random forest and SVM algorithms,the classification accuracy of the results obtained on the driving behavior dataset is improved by 6%and 8%,respectively.
作者
孙乾智
Sun Qianzhi(School of Computer Science,Southwest Petroleum University,Chengdu 610500,China)
出处
《无线互联科技》
2024年第2期88-93,97,共7页
Wireless Internet Technology
基金
西南石油大学自然科学青年基金,项目名称:高密度地震数据的初至波自动拾取方法研究,项目编号:41604114。
关键词
属性约简
SVM算法
随机森林算法
因子分析
驾驶行为
attribute reduction
SVM algorithm
random forest algorithm
factor analysis
driving behavior
