.png){kind=icon}
返回
An improved nonlinear correlation method for feature selection of complex data
.png){kind=icon}
PRE
AI
DOI：10.1007/s11071-023-08406-w.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
In this paper, the affine-invariant Gini distance correlation/covariance is put forward to characterize the dependence between categorical and numerical variables more effectively. Furtherly, the useful features from analyzing subjects are extracted for the establishment of a reliable discrimination model, and proper measurement of correlation is built. The affine-invariant Gini distance measures are not only orthogonal invariant, but also affine invariant, which is important when considering the preservation of the equivalent statistical inference. More importantly, the Gini distance measurements capture nonlinear dependence as well as independence, which is more superb than the existing methods. The affine-invariant Gini distance estimators are easy to calculate, and the estimation of probability density of variables is not required. Besides, it can be easily extended to the reproducing kernel Hilbert spaces. The simulation- and reality-based experiments illustrate that the improved nonlinear correlation method is more effective in measuring dependencies and selecting useful features when compared to other methods.
Keyword:
Complex data
Nonlinear correlation
Feature selection
Gini distance correlation
Affine invariant
期刊
IF:
6
论文数:
1.4W
被引数:
4.1W
机构
引用论文
Feature Redundancy Based on Interaction Information for Multi-Label Feature Selection.png)
IEEE ACCESS
OAAI
IEEE ACCESSIF3.6
OAAI分享收藏Cable force monitoring system of cable stayed bridges using accelerometers inside mobile smart phone
PREAI
分享收藏EQUIVALENCE OF DISTANCE-BASED AND RKHS-BASED STATISTICS IN HYPOTHESIS TESTINGANNALS OF STATISTICSOAAI
ANNALS OF STATISTICSIF3.7
OAAI分享收藏
查看更多内容