摘要
最小绝对值收缩和选择算子(LASSO)在运动想象脑电(EEG)特征选择中已得到了广泛应用。然而,LASSO的使用形式存在不少差异,其各种使用形式的性能如何目前还没有系统的研究。为此,探讨了现有的LASSO特征选择方法,进而提出了基于LASSO和主成分分析(PCA)降维的混合特征选择方法。该方法先训练LASSO模型,然后选择模型权重大于0的特征进行PCA降维,最后使用降维后的特征训练分类器。最优LASSO模型参数、特征个数、主成分个数使用10折交叉验证进行选择,分别使用Fisher线性判别分析(FLDA)、贝叶斯线性判别分析(BLDA)、支持向量机(SVM)3种分类器的分类结果作为交叉验证的评价准则。最优特征子集进行PCA降维之后,训练以上3种分类器作为最终的分类模型。选用两个公开的脑机接口竞赛数据验证算法的有效性,所提出的方法取得了80.06%的最高平均分类准确率。实验结果表明,所提出的方法优于现有的LASSO特征选择方法。
The least absolute shrinkage and selection operator(LASSO)has been widely used in motor imagery electroencephalogram(EEG)feature selection.However,there are many differences in the use forms of LASSO,and there is no systematic study on the performance of various forms.To this end,the existing LASSO feature selection methods are first systematically discussed,and then a hybrid feature selection method based on LASSO and principal component analysis(PCA)dimensionality reduction is proposed.In the proposed method,the LASSO model is first trained,then the features with model weights greater than 0 are selected for PCA dimensionality reduction,and finally,the reduced features are used for training the classifier.The optimal LASSO model parameters,number of features,and number of principal components are selected by 10-fold cross-validation,and the classification results of three classifiers,Fisher linear discriminant analysis(FLDA),Bayesian linear discriminant analysis(BLDA),and support vector machine(SVM),are used as the evaluation criteria of cross-validation.After PCA dimensionality reduction is performed on the optimal feature subset,the above three classifiers are trained as the final classification model.Two public brain-computer interface competition data are selected to verify the effectiveness of the algorithm,and the proposed method achieves the highest average classification accuracy of 80.06%.Experimental results show that the proposed method outperforms existing LASSO feature selection methods.
作者
莫云
梁国富
路仲伟
李智
许川佩
张绍荣
Mo Yun;Liang Guofu;Lu Zhongwei;Li Zhi;Xu Chuanpei;Zhang Shaorong(School of Electronic Information and Automation,Guilin University of Aerospace Technology,Guilin 541004,China;School of Electronic Engineering and Automation,Guilin University of Electronic Technology,Guilin 541004,China)
出处
《国外电子测量技术》
北大核心
2022年第5期9-14,共6页
Foreign Electronic Measurement Technology
基金
2020年广西高校中青年教师科研基础能力提升项目(2020KY21017)
桂林航天工业学院校级科研基金(XJ21KT27)项目资助
