摘要
提出一种基于免疫克隆多目标优化算法的特征选择方法,先将非监督特征选择问题归结为多目标优化问题,然后构造相应的问题模型和目标函数.最后,采用免疫克隆多目标优化算法,通过增加相关特征的显著性,减小不相关特征的显著性来实现每个特征显著性的优化,达到特征选择的目的.UCI数据集的仿真实验表明,该算法降低了错误识别率,验证了其在非监督特征选择中的应用潜力.
The unsupervised feature selection is transferred into a multiobjective optimization problem, and the immune clonal selection algorithm for multi-objective optimization is applied to solve it. Firstly, the unsupervised feature selection problem is translated into multi-objective problem. Secondly, the model and the objective functions are constructed. Lastly, each feature of significance is optimized by increasing the significance of the related features and decreasing the significance of the unrelated features. Experimental results on UCI data sets show that the error recognition rate is decreased and that the effectiveness and potential of the method are validated.
出处
《西安电子科技大学学报》
EI
CAS
CSCD
北大核心
2010年第1期18-22,共5页
Journal of Xidian University
基金
国家"863"计划资助项目(2009AA12Z210)
陕西省"13115"科技创新工程重大科技专项资助项目(2008ZDKG-37)
国家自然科学基金资助项目(60703107
60703108
60803098)
陕西省自然科学基金资助项目(2007F32)
国家教育部博士点基金资助项目(20070701022)
中国博士后科学基金特别资助项目(200801426)
中国博士后科学基金资助项目(20080431228)
教育部长江学者和创新团队支持计划资助项目(IRT0645)
关键词
非监督特征选择
克隆选择
多目标优化
unsupervised feature selection
clonal selection
multi-objective optimization