一种基于改进灰色关联分析的变量选择算法

A variable selection algorithm based on improved grey relational analysis

针对灰色绝对关联度模型和灰色相似关联度模型存在的问题,提出一种基于相对变化面积的改进灰色关联度模型.以序列几何形状的相似程度为基础,构建反应折线相似程度的相对变化面积,并以此作为关联系数的计算依据,同时以局部关联度的平均值度量整体的相似性,定义灰色关联度模型.此外,根据关联度计算结果,提出一种基于集合思想的变量选择算法,有效去除变量间的无关和冗余变量.仿真结果验证了所提出算法的有效性和合理性.

An improved grey relation model based on relative area change is proposed for the deficiency of the similitude degree of grey incidence and the absolute degree of grey incidence. Depending on the similitude degree of sequence curves, the relative area change that reflects the similitude degree of the sequence curves is constructed, and then the new relational coefficient is defined. At the same time, the new grey relation model is defined based on the mean value of the relational coefficients. Furthermore, according to the result of the relational analysis, a variable selection method is also proposed in order to reduce the irrelevant and redundant variables. The simulation results show the rationality and the effectiveness of the proposed algorithm.