一种拟合曲线的新拟合算法

A New Fitting Algorithm for Fitting Curve

非线性测量中对测量结果进行曲线拟合通常采用最小均方误差的标准进行系统参数的辨识,该方法未考虑输入样本的随机性。基于此,提出了一种新的拟合算法:考虑输入样本的随机性所采取的一种加权的最小均方误差拟合方法,利用已知的输入样本的统计特性对权值进行估计,对算法概率为1的任意逼近性给出证明,并对区间半长作出估计。实验表明,该算法有较理想的逼近效果。

The measurement data processed by mean square error is usually adopted in identifying curve fitting in nonlinear survey, the criterion of minimal system parameter. But sample randomness isn＇ t taken account of in this method. Therefore a new curve fitting algorithm is suggested in this paper： a weighted curve fitting method of minimal mean square error is adopted, in which the input sample is random, the weight is evaluated by statistical characteristic of input sample, intended approximation with probability equivalent to 1 is proved, and interval half-length is estimated. Experiment shows that the algorithm could work with better approximation effect.