实验二:逻辑回归算法实验

【实验目的】

理解逻辑回归算法原理,掌握逻辑回归算法框架;
理解逻辑回归的sigmoid函数;
理解逻辑回归的损失函数;
针对特定应用场景及数据,能应用逻辑回归算法解决实际分类问题。

【实验内容】

1.根据给定的数据集,编写python代码完成逻辑回归算法程序,实现如下功能:

建立一个逻辑回归模型来预测一个学生是否会被大学录取。假设您是大学部门的管理员,您想根据申请人的两次考试成绩来确定他们的入学机会。您有来自以前申请人的历史数据,可以用作逻辑回归的训练集。对于每个培训示例,都有申请人的两次考试成绩和录取决定。您的任务是建立一个分类模型,根据这两门考试的分数估计申请人被录取的概率。
算法步骤与要求:

(1)读取数据;(2)绘制数据观察数据分布情况;(3)编写sigmoid函数代码;(4)编写逻辑回归代价函数代码;(5)编写梯度函数代码;(6)编写寻找最优化参数代码(可使用scipy.opt.fmin_tnc()函数);(7)编写模型评估(预测)代码,输出预测准确率;(8)寻找决策边界,画出决策边界直线图。

\2. 针对iris数据集,应用sklearn库的逻辑回归算法进行类别预测。

要求:

(1)使用seaborn库进行数据可视化;(2)将iri数据集分为训练集和测试集(两者比例为8:2)进行三分类训练和预测;(3)输出分类结果的混淆矩阵。

【实验报告要求】

对照实验内容,撰写实验过程、算法及测试结果;
代码规范化:命名规则、注释;
实验报告中需要显示并说明涉及的数学原理公式;
查阅文献,讨论逻辑回归算法的应用场景;

1. 读取数据

import pandas as pd
Data=pd.read_csv("ex2data1.txt",header=None,names=['Exam1', 'Exam2', 'Admitted'])
Data
Exam1 Exam2 Admitted
0 34.623660 78.024693 0
1 30.286711 43.894998 0
2 35.847409 72.902198 0
3 60.182599 86.308552 1
4 79.032736 75.344376 1
5 45.083277 56.316372 0
6 61.106665 96.511426 1
7 75.024746 46.554014 1
8 76.098787 87.420570 1
9 84.432820 43.533393 1
10 95.861555 38.225278 0
11 75.013658 30.603263 0
12 82.307053 76.481963 1
13 69.364589 97.718692 1
14 39.538339 76.036811 0
15 53.971052 89.207350 1
16 69.070144 52.740470 1
17 67.946855 46.678574 0
18 70.661510 92.927138 1
19 76.978784 47.575964 1
20 67.372028 42.838438 0
21 89.676776 65.799366 1
22 50.534788 48.855812 0
23 34.212061 44.209529 0
24 77.924091 68.972360 1
25 62.271014 69.954458 1
26 80.190181 44.821629 1
27 93.114389 38.800670 0
28 61.830206 50.256108 0
29 38.785804 64.995681 0
... ... ... ...
70 32.722833 43.307173 0
71 64.039320 78.031688 1
72 72.346494 96.227593 1
73 60.457886 73.094998 1
74 58.840956 75.858448 1
75 99.827858 72.369252 1
76 47.264269 88.475865 1
77 50.458160 75.809860 1
78 60.455556 42.508409 0
79 82.226662 42.719879 0
80 88.913896 69.803789 1
81 94.834507 45.694307 1
82 67.319257 66.589353 1
83 57.238706 59.514282 1
84 80.366756 90.960148 1
85 68.468522 85.594307 1
86 42.075455 78.844786 0
87 75.477702 90.424539 1
88 78.635424 96.647427 1
89 52.348004 60.769505 0
90 94.094331 77.159105 1
91 90.448551 87.508792 1
92 55.482161 35.570703 0
93 74.492692 84.845137 1
94 89.845807 45.358284 1
95 83.489163 48.380286 1
96 42.261701 87.103851 1
97 99.315009 68.775409 1
98 55.340018 64.931938 1
99 74.775893 89.529813 1

100 rows × 3 columns

2. 数据分布状况

import matplotlib.pyplot as plt
#将录取和未录取进行分类
positive=Data[Data["Admitted"]==1]
negative=Data[Data["Admitted"]==0]
fig,ax=plt.subplots(figsize=(12,8))#创建画布,设置画布的大小
ax.scatter(positive['Exam1'],positive['Exam2'],s=30,c='g',marker='x',label='Admitted')#散点图,添加配置项
ax.scatter(negative['Exam1'],negative['Exam2'],s=30,c='r',marker='o',label='Not Admitted')
ax.legend() # 添加图例
ax.set_xlabel('Exam1 Score')
ax.set_ylabel('Exam2 Score')#添加x,y轴标签
Text(0,0.5,'Exam2 Score')

Free Photo Sharing by ThumbSnap

3. sigmoid函数

import numpy as np
def sigmoid(z):
    return 1/(1+np.exp(-z))
nums=np.arange(-10,10,step=1)
fig,ax=plt.subplots(figsize=(12,8))#设置画布的大小
ax.plot(nums,sigmoid(nums),"r")#在坐标轴上画线

Free Photo Sharing by ThumbSnap

4. logistics回归代价函数

def model(x,theta):
    return sigmoid(np.dot(x,theta.T))#dot矩阵的乘法运算 T转置

def cost(theta,x,y):
    theta = np.matrix(theta) #参数theta是一维数组,进行矩阵想乘时要把theta先转换为矩阵
    L1=np.multiply(-y,np.log(model(x,theta)))#multiply()数组和矩阵对应位置相乘
    L2=np.multiply(1-y,np.log(1-model(x,theta)))
    return np.sum(L1-L2)/(len(x))

Data.insert(0, 'Ones', 1)
cols=Data.shape[1]
x=np.array(Data.iloc[:,0:cols-1])#1-倒数第1列的数据
y=np.array(Data.iloc[:,cols-1:cols])#倒数第1列的数据
theta=np.zeros(x.shape[1])#1行三列的矩阵全部填充为0
print(cost(theta,x,y))
0.6931471805599453

5. 求梯度

def gradient(theta,x,y):
    theta = np.matrix(theta) #要先把theta转化为矩阵
    grad=np.dot(((model(x,theta)-y).T),x)/len(x)
    return np.array(grad).flatten()#因为下面寻找最优化参数的函数(opt.fmin_tnc())要求传入的gradient函返回值需要是一维数组,
                                   #因此需要利用flatten()将grad进行转换以下 
gradient(theta,x,y)
array([ -0.1       , -12.00921659, -11.26284221])

6. 搜索最优化参数

import scipy.optimize as opt
result = opt.fmin_tnc(func=cost, x0=theta, fprime=gradient, args=(x, y))
result
(array([-25.16131864,   0.20623159,   0.20147149]), 36, 0)

7. 预测准确率

def predict(theta,x):
    theta=np.matrix(theta)
    temp=sigmoid(x*theta.T)
    return [1 if x >= 0.5 else 0 for x in temp]
theta=result[0]
predictValues=predict(theta,x)
hypothesis=[1 if a==b else 0 for (a,b)in zip(predictValues,y)]
accuracy=hypothesis.count(1)/len(hypothesis)
print ('accuracy = {0}%'.format(accuracy*100))
accuracy = 89.0%

8. 得到决策边界并绘制

def find_x2(x1,theta):
    return [(-theta[0]-theta[1]*x_1)/theta[2] for x_1 in x1]
x1=np.linspace(30,100,1000)
x2=find_x2(x1,theta)
Data1=Data[Data['Admitted']==1]
Data2=Data[Data['Admitted']==0]
fig,ax=plt.subplots(figsize=(12,8))
ax.scatter(Data1['Exam1'],Data1['Exam2'],c='g',marker='x',label='Admitted')
ax.scatter(Data2['Exam2'],Data2['Exam1'],c='r',marker='o',label="Not Admitted")
ax.plot(x1,x2,'b',label="decision boundary")
ax.legend(loc=1)
ax.set_xlabel('Exam1 score')
ax.set_ylabel('Exam2 score')
ax.set_title("Training data with decision boundary")
plt.show()

Free Photo Sharing by ThumbSnap

9. 对iris数据集数据分析

import seaborn as sns
from sklearn.datasets import load_iris# 我们利用 sklearn 中自带的 iris 数据作为数据载入,并利用Pandas转化为DataFrame格式
data=load_iris() #得到数据特征
iris_target=data.target #得到数据对应的标签
iris_features=pd.DataFrame(data=data.data, columns=data.feature_names) #利用Pandas转化为DataFrame格式
iris_features.describe()
iris_all=iris_features.copy()
iris_all['target']=iris_target
#利用value_counts函数查看每个类别数量
pd.Series(iris_target).value_counts()
sns.pairplot(data=iris_all) # pairplot用来进行数据分析,画两两特征图。
plt.show()

Free Photo Sharing by ThumbSnap

10. 对iris数据集调用sk-learn的logistics回归进行类别预测

from sklearn.model_selection import train_test_split
X_train,X_test,y_train,y_test = train_test_split(iris_features,iris_target,test_size=0.3)
from sklearn.linear_model import LogisticRegression
clf=LogisticRegression(random_state=0,solver='lbfgs')
# 在训练集上训练逻辑回归模型
clf.fit(X_train,y_train)
print('the weight of Logistic Regression:\n',clf.coef_)
print('the intercept(w0) of Logistic Regression:\n',clf.intercept_)
train_predict=clf.predict(X_train)
test_predict=clf.predict(X_test)
the weight of Logistic Regression:
 [[-0.39310536  0.77102865 -2.124489   -0.91813939]
 [-0.28361972 -1.80748255  0.69900858 -1.23623074]
 [-0.32296799 -0.38952886  2.55652139  2.16024956]]
the intercept(w0) of Logistic Regression:
 [  6.20645518   5.06911911 -12.89262715]

11. 将iris数据集分0.7训练集0.3测试集进行三分类训练和预测

from sklearn.model_selection import train_test_split
X_train,X_test,y_train,y_test = train_test_split(iris_features,iris_target,test_size=0.3)
from sklearn.linear_model import LogisticRegression
clf=LogisticRegression(random_state=0,solver='lbfgs')
# 在训练集上训练逻辑回归模型
clf.fit(X_train,y_train)
print('the weight of Logistic Regression:\n',clf.coef_)
print('the intercept(w0) of Logistic Regression:\n',clf.intercept_)
train_predict=clf.predict(X_train)
test_predict=clf.predict(X_test)
the weight of Logistic Regression:
 [[-0.39310536  0.77102865 -2.124489   -0.91813939]
 [-0.28361972 -1.80748255  0.69900858 -1.23623074]
 [-0.32296799 -0.38952886  2.55652139  2.16024956]]
the intercept(w0) of Logistic Regression:
 [  6.20645518   5.06911911 -12.89262715]

12. 输出iris数据集分类结果的confusion matrix

from sklearn import metrics
#利用accuracy(准确度)【预测正确的样本数目占总预测样本数目的比例】评估模型效果
print('The accuracy of the Logistic Regression is:',metrics.accuracy_score(y_train,train_predict))
print('The accuracy of the Logistic Regression is:',metrics.accuracy_score(y_test,test_predict))
#查看混淆矩阵(预测值和真实值的各类情况统计矩阵)
confusion_matrix_result=metrics.confusion_matrix(y_test,test_predict)
print('The confusion matrix result:\n',confusion_matrix_result)
# 利用热力图对于结果进行可视化,画混淆矩阵
plt.figure(figsize=(8,6))
sns.heatmap(confusion_matrix_result,annot=True,cmap='Reds')
plt.xlabel('Predictedlabels')
plt.ylabel('Truelabels')
plt.show()
The accuracy of the Logistic Regression is: 0.9583333333333334
The accuracy of the Logistic Regression is: 0.8
The confusion matrix result:
 [[10  0  0]
 [ 0  7  3]
 [ 0  3  7]]

Free Photo Sharing by ThumbSnap

13. 算法涉及的数学原理

(1)sigmoid

\[y=f(x)=\frac{1}{1+e^{-x} } \]

(2)梯度下降

\[J(\theta)=\frac{1}{2}\left[h_{t}(x)-y\right]^{2} \]

\[\frac{\partial J}{\partial \theta_{j}}=-\frac{1}{m} \sum_{i=1}^{n}\left(y_{i}-h_{\theta}\left(x_{i}\right)\right) x_{i j} \]

\[\begin{array}{l} l(\theta)=\log L(\theta)=\sum_{i=1}\left(y_{i} \log h_{\theta}\left(x_{i}\right)+\left(1-y_{i}\right) \log \left(1-h_{\theta}\left(x_{i}\right)\right)\right) \\ \frac{\delta}{\delta_{\theta}} J(\theta)=-\frac{1}{m} \sum_{i=1}^{m}\left(y_{i} \frac{1}{h_{\theta}\left(x_{i}\right)} \frac{\delta}{\delta_{\theta}} h_{\theta}\left(x_{i}\right)-\left(1-\mathrm{y}_{\mathrm{i}}\right) \frac{1}{1-h_{\theta}\left(x_{i}\right)} \frac{\delta}{\delta_{\theta}} h_{\theta}\left(x_{i}\right)\right) \\ =-\frac{1}{m} \sum_{i=1}^{m}\left(y_{i} \frac{1}{g\left(\theta^{\mathrm{T}} x_{i}\right)}-\left(1-\mathrm{y}_{\mathrm{i}}\right) \frac{1}{1-g\left(\theta^{\mathrm{T}} x_{i}\right)}\right) \frac{\delta}{\delta_{\theta}} g\left(\theta^{\mathrm{T}} x_{i}\right) \\ =-\frac{1}{m} \sum_{i=1}^{m}\left(y_{i} \frac{1}{g\left(\theta^{\mathrm{T}} x_{i}\right)}-\left(1-\mathrm{y}_{\mathrm{i}}\right) \frac{1}{1-g\left(\theta^{\mathrm{T}} x_{i}\right)}\right) g\left(\theta^{\mathrm{T}} x_{i}\right)\left(1-g\left(\theta^{\mathrm{T}} x_{i}\right)\right) \frac{\delta}{\delta_{\theta}} \theta^{\mathrm{T}} x_{i} \\ =-\frac{1}{m} \sum_{i=1}^{m}\left(y_{i}\left(1-g\left(\theta^{\mathrm{T}} x_{i}\right)\right)-\left(1-\mathrm{y}_{\mathrm{i}}\right) g\left(\theta^{\mathrm{T}} x_{i}\right)\right) x_{i}^{j} \\ =-\frac{1}{m} \sum_{i=1}^{m}\left(y_{i}-g\left(\theta^{\mathrm{T}} x_{i}\right)\right) x_{i}^{j} \\ =\frac{1}{m} \sum_{i=1}^{m}\left(h_{\theta}\left(x_{i}\right)-y_{i}\right) x_{i}^{j} \end{array} \]

\[\theta_{n+1}=\theta_{n}-\alpha J^{\prime}(\theta) \]

(3)损失函数

\[J(\theta)=\frac{1}{m} \sum_{i=1}^{m} \operatorname{Cos} t\left(h_{\theta}\left(x^{(i)}\right), y^{(i)}\right) \]

\[J(\theta)=\frac{1}{m} \sum_{i=1}^{m} \operatorname{Cos} t\left(h_{\theta}\left(x^{(i)}\right), y^{(i)}\right)\operatorname{Cost}\left(h_{\theta}(x), y\right)=\left\{\begin{aligned} -\log \left(h_{\theta}(x)\right) & \text { if } y=1 \\ -\log \left(1-h_{\theta}(x)\right) & \text { if } y=0 \end{aligned}\right. \]

14. 逻辑回归算法的应用场景

  1. 用于分类场景, 尤其是因变量是二分类(0/1,True/False,Yes/No)时我们应该使用逻辑回归。
  2. 不要求自变量和因变量是线性关系

15. 存在的问题

  1. 防止过拟合和低拟合,应该让模型构建的变量是显著的。一个好的方法是使用逐步回归方法去进行逻辑回归。
  2. 逻辑回归需要大样本量,因为最大似然估计在低样本量的情况下不如最小二乘法有效。
  3. 独立的变量要求没有共线性。
posted @ 2022-10-31 09:14  qsl0000  阅读(102)  评论(0编辑  收藏  举报