梯度下降法求解线性回归问题 python

（1）编码实现基于梯度下降的单变量和多变量线性回归算法，包括梯度的计算与验证；

（2）画数据散点图和求得的直线；

（3）画梯度下降过程中损失的变化图；

(4）基于训练得到的参数，输入新的样本数据，输出预测值；

import numpy as np
from matplotlib import pyplot as plt
train_data = np.genfromtxt('C:/Users/Administrator/Desktop/data.csv',delimiter=',')
x_data = train_data[:,0]
y_data = train_data[:,1]

lr = 0.0001
theat0 = 0
theat1 = 0
iteators = 200

def compute_loss(x_data, y_data,theat0,theat1 ):
    sum = 0

    for i in range(len(x_data)):
        sum += pow((theat0+theat1*x_data[i])-y_data[i],2)

    return sum/(2*len(x_data))

def gradient_fun(x_data, y_data,theat0,theat1,iteators,lr):
    m = float(len(x_data))
    
    cost = np.zeros(iteators)

    for j in range(iteators):
        sum1 = 0
        sum2 = 0
        for i in range(len(x_data)):
            sum1 += (theat0+theat1*x_data[i])-y_data[i]
            sum2 += ((theat0+theat1*x_data[i])-y_data[i])*x_data[i]

        theat0 = theat0 - lr*sum1/m
        theat1 = theat1 - lr*sum2/m
        cost[j] =compute_loss(x_data,y_data,theat0,theat1)

        print('iteators = ',j,'loss =',compute_loss(x_data,y_data,theat0,theat1),'theat0 = ',theat0,'theat1 = ',theat1)

    return theat0,theat1,cost

theat0,theat1,cost= gradient_fun(x_data, y_data,theat0,theat1,iteators,lr)

plt.plot(x_data,y_data,'b.')
plt.plot(x_data,theat0+theat1*x_data,'r')
plt.show()

fig,ax =plt.subplots(figsize=(12,8))
ax.plot(np.arange(iteators),cost,'b')
ax.set_xlabel('Iterations')
ax.set_ylabel('Cost')
ax.set_title('cost__iteators')
plt.show()

pre_data = np.genfromtxt('C:/Users/Administrator/Desktop/data.csv',delimiter=',')
x_data1 =pre_data[:,0]
def predict_func(x_data1,theat0,theat1):
    theat0 = 0.03501490948451012
    theat1 = 1.4788038980136102
    pre_y = np.zeros(len(x_data1))
    for i in range(len(x_data1)):
        pre_y[i] = theat0+theat1*x_data[i]
        print("x  ",x_data[i],"预测值",pre_y[i])
theat0 =  0.03501490948451012
theat1 =  1.4788038980136102  
predict_func(x_data1,theat0,theat1)

jupyter上运行，运行时自己调整data路径
data文件找不到了，换成你们自己的也可以呜呜呜。