基于核函数的稀疏属性选择算法

鉴于传统属性选择算法无法捕捉属性之间的关系的问题,文中提出了一种非线性属性选择方法。该方法通过引入核函数,将原始数据集投影到高维的核空间,因在核空间内进行运算,进而可以考虑到数据属性之间的关系。由于核函数自身的优越性,即使数据通过高斯核投影到无穷维的空间中,计算复杂度亦可以控制得较小。在正则化因子的限制上,使用两种范数进行双重约束,不仅提高了算法的准确率,而且使得算法实验结果的方差仅为0.74,远小于其他同类对比算法,且算法更加稳定。在8个常用的数据集上将所提算法与6个同类算法进行比较,并用SVM分类器来测试分类准确率,最终该算法得到最少1.84%,最高3.27%,平均2.75%的提升。

Sparse Feature Selection Algorithm Based on Kernel Function

In view of the condition that the traditional feature selection algorithm can not capture the relationship between features,a nonlinear feature selection method was proposed.By introducing a kernel function,the method projects the original data set into a high-dimensional kernel space,and considers the relationship between sample features by performing operations in the kernel space.Due to the superiority of the kernel function,even if the data are projected into the infinite dimensional space through the Gaussian kernel,the computational complexity can be controlled to a small extent.For the limitation of the regularization factor,the use of two norms for double constraint not only improves the accuracy of the algorithm,but also makes the variance of the algorithm only be 0.74,which is much smaller than other similar comparison algorithms,and it is more stable.6 similar algorithms were compared on 8 common data sets,and the SVM classifier was used to test the effect.The results demonstrate that the proposed algorithm can get the improvement by a minimum of 1.84%,a maximum of 3.27%,and an average of 2.75%.