Week1 Python学习之线性回归

#本文中所使用的数据集参考吴恩达的course课程题目，数据集下载地址：链接：百度网盘 请输入提取码 密码：mwcl

1 加载数据集并可视化

使用numpy科学计算库加载数据集中数据，并通过matplotlib库下plot函数进行绘制散点图。

import numpy as np
import matplotlib.pyplot as plt
from matplotlib.font_manager import FontProperties

data = np.loadtxt('D:\ZacLing\mat\Study\ex1data1.txt',delimiter = ',')
x = data[:,0]
y = data[:,1]
plt.figure()
plt.plot(x,y,'k.')
plt.show()

2 sklearn版本

线性式（假设函数）：

损失函数：

2.1 初始化参数

Len = np.size(X)
X = np.hstack(np.ones([Len,1]),X.reshape(Len,1))#增加常数项

2.2 初始化模型

from sklearn import linear_model
model = linear_model.LinearRegression()

2.3 拟合数据并画图

model.fit(X,y)
xp = np.array([[1,0],[1,3],[1,7],[1,12],[1,15],[1,25]])
yp = model.predict(xp)
print(yp)
plt.plot(xp[:,1],yp,'g-')
plt.show()

 2.4 程序总览

import numpy as np
import matplotlib.pyplot as plt
from matplotlib.font_manager import FontProperties
from sklearn import linear_model

data = np.loadtxt('ex1data1.txt',delimiter = ',')
X = data[:,0]
y = data[:,1]
plt.figure()
plt.plot(X,y,'r.')
Len = np.size(X)
X.shape = (Len,1)
X = np.hstack([np.ones([Len,1]),X.reshape(Len,1)]) #增加常数项

model = linear_model.LinearRegression()
model.fit(X,y)
xp = np.array([[1,0],[1,3],[1,7],[1,12],[1,15],[1,25]])
yp = model.predict(xp)
print(yp)
plt.plot(xp[:,1],yp,'g-')
plt.show()

3 手撸版本

3.1 损失函数图像绘制

import numpy as np
from mpl_toolkits import mplot3d
import matplotlib.pyplot as plt
from matplotlib.font_manager import FontProperties

def CostFunction(X,y,theta0,theta1): #损失函数
    n = np.size(theta0)
    J_list = np.zeros([n,n])
    m = np.size(y)
    for i in range(0,n):
        for j in range(0,n):
            theta = np.array([[theta0[i]],[theta1[j]]])
            J_list[i,j] = sum(np.square(np.dot(X,theta) - y)) / (2 * m)
    return J_list

data = np.loadtxt('D:\ZacLing\mat\Study\ex1data1.txt',delimiter = ',')
X = data[:,0]
y = data[:,1]
plt.figure()
ax = plt.axes(projection = '3d')
Len = np.size(X)
X.shape = (Len,1)
y.shape = (Len,1)
X = np.hstack([np.ones([Len,1]),X.reshape(Len,1)]) #增加常数项
theta0 = np.linspace(-6,0,200)
theta1 = np.linspace(1,1.5,200)
ax.contour3D(theta0,theta1,CostFunction(X,y,theta0,theta1),50, cmap='cool')
plt.show()

3.2 程序总览

import numpy as np
import matplotlib.pyplot as plt
from matplotlib.font_manager import FontProperties

def CostFunction(X,y,theta): #损失函数
    m = np.size(y)
    J = 0
    J = sum(np.square(np.dot(X,theta) - y)) / (2 * m)
    return J

def GradientDescent(X,y,theta,alpha,steps): #梯度下降
    m = np.size(y)
    J_list = np.zeros([steps,1])
    for i in range(0,steps):
        h = np.dot(X,theta)
        temp = np.zeros([2,1])
        for j in range(0,m):
            temp = temp + (np.dot((h[j] - y[j]).reshape(1,1),X[j,:].reshape(1,2))).reshape(2,1)
        theta = theta - (alpha * temp) / m
        J_list[i] = CostFunction(X,y,theta)
    return theta

data = np.loadtxt('ex1data1.txt',delimiter = ',')
X = data[:,0]
y = data[:,1]
plt.figure()
plt.plot(X,y,'r.')
Len = np.size(X)
X.shape = (Len,1)
y.shape = (Len,1)
X = np.hstack([np.ones([Len,1]),X.reshape(Len,1)]) #增加常数项
theta = np.zeros([2,1]) #初始化系数向量
steps = 1500 #迭代次数
alpha = 0.01 #学习率
theta = GradientDescent(X,y,theta,alpha,steps)
xp = np.array([[1,0],[1,3],[1,7],[1,12],[1,15],[1,25]])
yp = np.dot(xp,theta)
plt.plot(xp[:,1],yp,'g-')
print(theta)
plt.show()

4 多元线性回归

数据集下载地址：百度网盘 请输入提取码 密码：29fv

4.1 数据归一化

def Normalization(X):
    range_ = np.max(X) - np.min(X)
    return (X - np.min(X)) / range_

4.2 程序总览

from unicodedata import normalize
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.font_manager import FontProperties
from sklearn import linear_model

def Normalization(X):
    range_ = np.max(X) - np.min(X)
    return (X - np.min(X)) / range_

data = np.loadtxt('ex1data2.txt',delimiter = ',')
X = data[:,0:2]
y = data[:,2]
X = Normalization(X)
model = linear_model.LinearRegression()
model.fit(X,y)
print(model.intercept_)
print(model.coef_)