基于split-and-conquer和非参数向前选择法的变量选择

变量选择是统计学界研究的重要课题之一.当处理高维数据时,一些常用的变量选择方法大多比较耗时,因此提出了一种既能筛选出高维数据中的变量又能节省时间的方法:基于split-and-conquer的非参数向前选择法.首先使用split-and-conquer方法将数据进行拆分,然后使用B样条函数逼近的非参数向前选择法进行研究.实验结果表明:基于split-and-conquer的非参数向前选择法可以较好地将变量选择出来,并且节省了大量时间.

Non-parametric forward selection method based on split-and-conquer

Variable selection is one of the important topics in the field of statistics. When dealing with high-dimensional data, most of the commonly used variable selection methods are time-consuming. This paper proposes a method that can filter out the variables in high-dimensional data and save time, that is, the non-parametric forward selection method based on split-and-conquer. Firstly, the split-and-conquer method is used to split the data, and then the non-parametric forward selection method of B-spline function approximation is used. The experimental results show that the non-parametric forward selection method based on split-and-conquer can select variables well, and save a lot of time.