摘要
针对核密度估计中窗宽确定困难的问题,提出了基于估计点的滑动变窗宽核密度估计算法。通过采用固定窗宽的密度估计函数代替假设的正态分布密度函数的方法,对估计域中的每一估计点求取其最优窗宽值,实现了窗宽根据样本的分布情况,在不同的估计点自动调整窗宽的取值。文中给出了算法的具体推导以及实现步骤,给出了对比试验。结果表明,新算法在估计结果的精度、平滑度等方面比现有固定窗宽的算法有明显提高。
Taking aim at that the bandwidth of the kernel density estimate is difficult to choose,a slide variable bandwidth kernel density estimation algorithm based on estimate point was proposed.The optimum bandwidth at every estimate point in the estimate interval was solved by the kernel density estimation function obtained with fixed bandwidth instead of the hypothetic normal one to adjust the bandwidth value automatically according to the sample distribution.The detailed deducing process,realizing steps and contrast experiments of the algorithm were put forward.The results show that the new algorithm has better precision and gliding property of kernel density estimate than that of the fixed bandwidth algorithm.
出处
《兵工学报》
EI
CAS
CSCD
北大核心
2009年第2期231-235,共5页
Acta Armamentarii
关键词
计算机应用
核密度估计
窗宽
估计点
算法
computer application
kernel density estimation
bandwidth
estimate point
algorithm
