非线性回归的一种算法

A new method of nonlinear regression

非线性回归一般有可化为线性回归与纯非线性回归２种情况。传统的解决方法是最小二乘法及Ｇａｕｓｓ－Ｎｅｗｔｏｎ迭代法。文章用有理插值函数去逼近非线性函数便可得到一近似的回归函数，计算结果表明，该方法拟合的平均绝对误差及均方差比传统方法效果好。

To deal with the two kinds of nonlinear regressions one can be changed into linear regression and the other is pure nonlinear regression,least square method and Gauss Newton interative method are the two conventional ways.A new method of using rational interpolation function to approach nonlinear function is proposed in this article. The mean absolute error and the mean square deviation obtained by the method show that the new method is more effective than the traditional iterative methods.