摘要
针对在数据量大且分布不均匀时非线性拟合效果不佳且易受离群点影响的问题,提出一种全局搜索密度峰值聚类中心的线性回归方法,采用密度峰值聚类方法根据样本分布密度将所有数据点划分为多个类簇,引入DKC值和AM度量降低离群点对聚类中心选取的影响,将聚类中心作为拟合的特征点进行分段线性拟合。通过对航班油耗的分段线性拟合的实验验证了该方法在此类拟合问题中的有效性,为线性回归分析提供了一种新思路。
Aiming at the problem of poor nonlinear fitting effect and susceptible to outliers when the amount of data is large and the distribution is uneven,a linear regression method of global search for the density peak clustering center is proposed,and the density peak clustering method is used according to the sample.The distribution density divides all data points into multiple clus-ters,introduces the DKC value and AM metric to reduce the influence of outliers on the selection of cluster centers,and uses the cluster centers as the fitted feature points for piecewise linear fitting.The experiment of piecewise linear fitting of flight fuel con-sumption verifies the effectiveness of this method in this type of fitting problem,and provides a new idea for linear regression analy-sis.
作者
马翔
MA Xiang(School of Electronics and Automation,Civil Aviation University of China,Tianjin 300300)
出处
《计算机与数字工程》
2024年第5期1353-1358,共6页
Computer & Digital Engineering
基金
国家科技支撑计划(编号:2012BAC20B0304)
民航局专项项目(编号:GH201661279)资助。
