数据拟合的最小二乘法

求其最小二乘逼近，取权系数为1，基函数：
${\phi }_1(x)=1,{\phi }_2(x)=x,{\phi }_3(x)=x^2,{\phi }_4(x)=x^3$
$$\begin{array}{c}{\Phi }_1=(1,1,1,\dots ,1,1{)}^T\\ {\Phi }_2=(-1,-0.75,-0.5,-0.25,0,0.25,0.5,0.75,1,1.25{)}^T\\ {\Phi }_3={\left((-1{)}^2,(-0.75{)}^2,(-0.5{)}^2,(-0.25{)}^2,0^2,0.25^2,0.5^2,0.75^2,1^2,1.25^2\right)}^T\\ {\Phi }_4={\left(-1^3,-0.75^3,-0.5^3,-0.25^3,0^3,0.25^3,0.5^3,0.75^3,1^3,1.25^3\right)}^T\\ Y=(0.2209,0.3295,0.8826,1.4392,2.0003,2.5645,3.1334,3.7061,4.2836,3.1334{)}^T\end{array}$$
一次多项式拟合数据：

$$拟合函数:f(x)={\partial }_1+{\partial }_{2}x$$
二次多项式拟合数据：

$$拟合函数:f(x)={\partial }_1+{\partial }_{2}x+{\partial }_3{x}^2$$
三次多项式拟合数据：

$$拟合函数:f(x)={\partial }_1+{\partial }_{2}x+{\partial }_3{x}^2+{\partial }_4{x}^3$$
$$误差计算:\sum _{n=1}^{10}{\left[Y_n-f\left(x_n\right)\right]}^2$$
程序求解：

import numpy as np

A1 = np.array([[1], [1], [1], [1], [1], [1], [1], [1], [1], [1]])
A2 = np.array([[-1], [-0.75], [-0.5], [-0.25], [0], [0.25], [0.5], [0.75], [1], [1.25]])
A3 = np.array([[(-1)**2], [(-0.75)**2], [(-0.5)**2], [(-0.25)**2], [0**2], [0.25**2], [0.5**2], [0.75**2], [1**2], [1.25**2]])
A4 = np.array([[-1**3], [-0.75**3], [-0.5**3], [-0.25**3], [0**3], [0.25**3], [0.5**3], [0.75**3], [1**3], [1.25**3]])
B = np.array([[0.2209], [0.3295], [0.8826], [1.4392], [2.0003], [2.5645], [3.1334], [3.7061], [4.2836], [3.1334]])

# 一次
a1 = np.array([[np.dot(A1.T, A1), np.dot(A1.T, A2)],
               [np.dot(A2.T, A1), np.dot(A2.T, A2)]]).reshape(2,2)
b1 = np.array([[np.dot(B.T, A1)], [np.dot(B.T, A2)]]).reshape(2,1)
c1 = np.linalg.solve(a1, b1)
print(c1) #拟合系数

C1 = c1[0] + c1[1] * A1
D1 = (B-C1)**2
print(np.sum(D1)) #误差

# 二次
a2 = np.array([[np.dot(A1.T, A1), np.dot(A1.T, A2), np.dot(A1.T, A3)],
               [np.dot(A2.T, A1), np.dot(A2.T, A2), np.dot(A2.T, A3)],
               [np.dot(A3.T, A1), np.dot(A3.T, A2), np.dot(A3.T, A3)]]).reshape(3,3)
b2 = np.array([[np.dot(B.T, A1)], [np.dot(B.T, A2)], [np.dot(B.T, A3)]]).reshape(3,1)
c2 = np.linalg.solve(a2, b2)
print(c2) #拟合系数

C2 = c2[0] + c2[1] * A1 + c2[2] * A1**2
D2 = (B-C2)**2
print(np.sum(D2)) #误差

# 三次
a3 = np.array([[np.dot(A1.T, A1), np.dot(A1.T, A2), np.dot(A1.T, A3), np.dot(A1.T, A4)],
               [np.dot(A2.T, A1), np.dot(A2.T, A2), np.dot(A2.T, A3), np.dot(A2.T, A4)],
               [np.dot(A3.T, A1), np.dot(A3.T, A2), np.dot(A3.T, A3), np.dot(A3.T, A4)],
               [np.dot(A4.T, A1), np.dot(A4.T, A2), np.dot(A4.T, A3), np.dot(A4.T, A4)]]).reshape(4,4)
b3 = np.array([[np.dot(B.T, A1)], [np.dot(B.T, A2)], [np.dot(B.T, A3)], [np.dot(B.T, A4)]]).reshape(4,1)
c3 = np.linalg.solve(a3, b3)
print(c3) #拟合系数

C3 = c3[0] + c3[1] * A1 + c3[2] * A1**2 + c3[3] * A1**3
D3 = (B-C3)**2
print(np.sum(D3)) #误差

从多项式拟合数据的误差平方和来看，二次多项式的拟合效果相对较好，其次是一次多项式拟合，三次多项式拟合效果相对较差。

如有错误或者不足之处，欢迎大家留言指正！