最小冗余最大分离准则特征选择方法
发布时间:2019-02-27 08:45
【摘要】:特征选择是处理高维数据的一项有效技术。针对传统方法的不足,结合F-score与互信息,提出了一种最小冗余最大分离的特征选择评价准则,该准则使所选择的特征具有更好的分类和预测能力;采用二进制布谷鸟搜索算法和二次规划两种搜索策略来搜索最优特征子集,并对两种搜索策略的准确性和计算量进行分析比较;最后,利用UCI数据集进行实验测试,实验结果说明了所提理论的有效性。
[Abstract]:Feature selection is an effective technique for processing high dimensional data. Aiming at the deficiency of traditional method, combining F-score and mutual information, this paper proposes a feature selection evaluation criterion of minimum redundancy and maximum separation, which makes the selected features have better classification and prediction ability. The binary cuckoo search algorithm and quadratic programming search strategy are used to search the optimal feature subset, and the accuracy and calculation amount of the two search strategies are analyzed and compared. Finally, the experimental results show that the proposed theory is effective by using UCI data set.
【作者单位】: 西安工程大学理学院;
【基金】:陕西省软科学研究项目(No.2014KRM28-01) 陕西省教育厅专项科研计划项目(No.16JK1341) 西安市2015基础教育研究重大招标项目(No.2015ZB-ZY04) 西安工程大学研究生创新基金资助项目(No.CX201614)
【分类号】:O212;TP18
本文编号:2431297
[Abstract]:Feature selection is an effective technique for processing high dimensional data. Aiming at the deficiency of traditional method, combining F-score and mutual information, this paper proposes a feature selection evaluation criterion of minimum redundancy and maximum separation, which makes the selected features have better classification and prediction ability. The binary cuckoo search algorithm and quadratic programming search strategy are used to search the optimal feature subset, and the accuracy and calculation amount of the two search strategies are analyzed and compared. Finally, the experimental results show that the proposed theory is effective by using UCI data set.
【作者单位】: 西安工程大学理学院;
【基金】:陕西省软科学研究项目(No.2014KRM28-01) 陕西省教育厅专项科研计划项目(No.16JK1341) 西安市2015基础教育研究重大招标项目(No.2015ZB-ZY04) 西安工程大学研究生创新基金资助项目(No.CX201614)
【分类号】:O212;TP18
【相似文献】
相关期刊论文 前4条
1 刘扬;刘伟江;;特征选择方法在信用评估指标选取中的应用[J];数理统计与管理;2006年06期
2 詹艳梅,曾向阳,孙进才;基于粗糙集理论的目标特征选择方法[J];自然科学进展;2004年12期
3 宋枫溪,高秀梅,刘树海,杨静宇;统计模式识别中的维数削减与低损降维[J];计算机学报;2005年11期
4 ;[J];;年期
,本文编号:2431297
本文链接:https://www.wllwen.com/kejilunwen/yysx/2431297.html