摘要
特征选择是模式识别和机器学习领域的重要问题。针对目前Filter和Wrapper方法,以及传统二阶段组合式方法存在的缺陷,提出了一种双重过滤式特征选择方法FSTPF,并在三个国际公认数据集和一个盾构隧道施工实时数据集上进行了验证测试。实验结果表明,FSTPF算法降维效果好,且获得的优化特征子集的分类准确率得到了提高。
Feature selection is an important problem in the pattern recognition and machine learning areas.Aimed at the question that there are some shortcomings in the actual Filter,Wrapper and tradictional two-phase combined methods,this paper proposes a Feature Selection algorithm based on Two-Phase Filter(FSTPF),and it is used to test in three international accepted datasets and a shield tunneling construncting real-time dataset.The emulational experiment shows that FSTPF can get good effect of reducting dimension and improve the classification accuracy of best feature subset.
出处
《计算机工程与应用》
CSCD
北大核心
2011年第19期190-193,206,共5页
Computer Engineering and Applications
基金
国家自然科学基金No.50778109
上海市重点学科项目(No.J50103)
浙江省林业厅面上项目(No.07A14)
浙江省自然科学基金项目(No.Y1080777)~~
关键词
特征选择
免疫
克隆选择算法
过滤
feature selection
immune
clone selection alogrithm
filter