python最小二乘法拟合直线

最小二乘法的推导可参考下面这个博客,推导非常详细.
https://my.oschina.net/keyven/blog/526010
首先画一个加入了噪声的散点图,函数是y=0.2x,加入噪声后如下图所示:

代码如下:

# -*- coding: utf-8 -*
import numpy as np
import matplotlib.pyplot as plt
x = np.linspace(0, 30, num=50)
y = 0.2*x+[np.random.random() for _ in range(50)]
if __name__ == '__main__':
    plt.figure(figsize=(10, 5), facecolor='w')
    plt.plot(x, y, 'ro', lw=2, markersize=6)
    plt.grid(b=True, ls=':')
    plt.xlabel(u'X', fontsize=16)
    plt.ylabel(u'Y', fontsize=16)
    plt.show()

我将上面这个代码保存为paint_scatter_plot.py这个名字,方便其他地方调用.
接下来画出拟合直线:

代码如下:

from  paint_scatter_plot import x,y
import numpy as np 
import matplotlib.pyplot as plt
def Least_squares(x,y):
    x_ = x.mean()
    y_ = y.mean()
    m = np.zeros(1)
    n = np.zeros(1)
    k = np.zeros(1)
    p = np.zeros(1)
    for i in np.arange(50):
        k = (x[i]-x_)* (y[i]-y_)
        m += k
        p = np.square( x[i]-x_ )
        n = n + p
    a = m/n
    b = y_ - a* x_
    return a,b
if __name__ == '__main__':
    a,b = Least_squares(x,y)
    print a,b
    y1 = a * x + b
    plt.figure(figsize=(10, 5), facecolor='w')
    plt.plot(x, y, 'ro', lw=2, markersize=6)
    plt.plot(x, y1, 'r-', lw=2, markersize=6)
    plt.grid(b=True, ls=':')
    plt.xlabel(u'X', fontsize=16)
    plt.ylabel(u'Y', fontsize=16)
    plt.show()

代码运算结果: a = 0.19784577, b = 0.48503344 ,还是非常接近正确结果的.