天天看点
返回

使用线性回归、逻辑回归、决策树、随机森林进行泰坦尼克救援预测泰坦尼克救援预测第一步：数据分析第二步：初步建模调整参数第三步：使用集成学习算法，随机森林

泰坦尼克救援预测 

from IPython.display import Image
Image(filename=r'C:\Users\a\Desktop\暑假\Titantic\QQ截图20190827081938.png',width=800)
使用线性回归、逻辑回归、决策树、随机森林进行泰坦尼克救援预测泰坦尼克救援预测第一步：数据分析第二步：初步建模调整参数第三步：使用集成学习算法，随机森林

第一步：数据分析 

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
           
#读取文件 查看数据描述
data = pd.read_csv('titanic_train.csv')
data.describe()
PassengerId Survived Pclass Age SibSp Parch Fare
count 891.000000 891.000000 891.000000 714.000000 891.000000 891.000000 891.000000
mean 446.000000 0.383838 2.308642 29.699118 0.523008 0.381594 32.204208
std 257.353842 0.486592 0.836071 14.526497 1.102743 0.806057 49.693429
min 1.000000 0.000000 1.000000 0.420000 0.000000 0.000000 0.000000
25% 223.500000 0.000000 2.000000 20.125000 0.000000 0.000000 7.910400
50% 446.000000 0.000000 3.000000 28.000000 0.000000 0.000000 14.454200
75% 668.500000 1.000000 3.000000 38.000000 1.000000 0.000000 31.000000
max 891.000000 1.000000 3.000000 80.000000 8.000000 6.000000 512.329200

可以看到数据中的Age列和Cabin列以及Embarked的列有缺失值，Cabin列缺失值数量太多，直接舍去，然后Ticket列对于实际的获救应该也没有什么关系

PassengerId Survived Pclass Name Sex Age SibSp Parch Ticket Fare Cabin Embarked
1 3 Braund, Mr. Owen Harris male 22.0 1 A/5 21171 7.2500 NaN S
1 2 1 1 Cumings, Mrs. John Bradley (Florence Briggs Th... female 38.0 1 PC 17599 71.2833 C85 C
2 3 1 3 Heikkinen, Miss. Laina female 26.0 STON/O2. 3101282 7.9250 NaN S
3 4 1 1 Futrelle, Mrs. Jacques Heath (Lily May Peel) female 35.0 1 113803 53.1000 C123 S
4 5 3 Allen, Mr. William Henry male 35.0 373450 8.0500 NaN S

查阅当时的背景资料发现在当时泰坦尼克失事之后的逃生政策是妇女和儿童优先，下面查看一下妇女和儿童的逃生率，可以看到女性的平均获救人数是0.74远大于男性的

#查阅当时的背景资料发现在当时泰坦尼克失事之后的逃生政策是妇女和儿童优先，下面查看一下妇女和儿童的逃生率
data.pivot_table(index='Sex',values='Survived')  #可以看到女性的平均获救人数是0.74远大于男性的0.18
Survived
Sex
female 0.742038
male 0.188908

查看获救人群的平均年龄 然而好像说明不了什么问题

#查看获救人群的平均年龄   然而好像说明不了什么问题
data.pivot_table(index='Survived',values='Age')
Age
Survived
30.626179
1 28.343690

查看女性人群的平均年龄 然而好像说明不了什么问题

#查看女性人群的平均年龄  然而好像说明不了什么问题
data[data['Sex']=='female'].pivot_table(index='Survived',values='Age')
Age
Survived
25.046875
1 28.847716

查阅资料发现当时儿童的定义为14岁以下，查看儿童的获救率 发现随着年龄的增长获救率会降低这也印证了妇女和儿童优先的逃生政策

#查阅资料发现当时儿童的定义为14岁以下，查看儿童的获救率  发现随着年龄的增长获救率会降低这也印证了妇女和儿童优先的政策  
#所以我们认为年龄是一个重要特征
for i in np.arange(20):
    print(data[data['Age']<= i].pivot_table(index = 'Sex',values='Survived'))
           
Empty DataFrame
Columns: []
Index: []
        Survived
Sex             
female       1.0
male         0.8
        Survived
Sex             
female  0.600000
male    0.642857
        Survived
Sex             
female  0.583333
male    0.722222
        Survived
Sex             
female  0.705882
male    0.652174
        Survived
Sex             
female  0.761905
male    0.652174
        Survived
Sex             
female  0.739130
male    0.666667
        Survived
Sex             
female  0.750000
male    0.615385
        Survived
Sex             
female  0.730769
male    0.607143
        Survived
Sex             
female  0.633333
male    0.593750
        Survived
Sex             
female  0.612903
male    0.575758
        Survived
Sex             
female  0.593750
male    0.555556
        Survived
Sex             
female  0.593750
male    0.567568
        Survived
Sex             
female  0.617647
male    0.567568
        Survived
Sex             
female  0.631579
male    0.538462
        Survived
Sex             
female  0.651163
male    0.525000
        Survived
Sex             
female  0.673469
male    0.431373
        Survived
Sex             
female  0.690909
male    0.396552
        Survived
Sex             
female  0.676471
male    0.338028
        Survived
Sex             
female  0.706667
male    0.292135

在当时的社会等级制度严格 查看一下三个船舱等级对应的获救率 发现船舱等级不同获救率也会有很大的不同所以船舱等级也是一个重要特征

#在当时的社会等级制度严格  查看一下三个船舱等级对应的获救率   发现船舱等级不同获救率也会有很大的不同所以船舱等级也是一个重要特征
data.pivot_table(index='Pclass',values='Survived')
Survived
Pclass
1 0.629630
2 0.472826
3 0.242363

那么登船地点会不会影响获救率呢 看起来登船地点对应的获救率也有较大区别，可能不同的登船地点上到船上的位置不同，距离逃生地点的远近也不同

#那么登船地点会不会影响获救率呢
data.pivot_table(index='Embarked',values='Survived')
#看起来登船地点对应的获救率也有较大区别，可能不同的登船地点上到船上的位置不同，距离逃生地点的远近也不同
Survived
Embarked
C 0.553571
Q 0.389610
S 0.336957

那么家里的兄弟姐妹的数量会不会影响获救率呢 可以看到越多的兄弟姐妹获救率越低

#那么家里的兄弟姐妹的数量会不会影响获救率呢  可以看到越多的兄弟姐妹获救率越低
data.pivot_table(index='SibSp',values='Survived')
Survived
SibSp
0.345395
1 0.535885
2 0.464286
3 0.250000
4 0.166667
5 0.000000
8 0.000000

那么家里老人和小孩的数量会不会影响获救率呢 可以看到总体来说老人和小孩的数量越多获救率也越大

#那么家里老人和小孩的数量会不会影响获救率呢     可以看到总体来说老人和小孩的数量越多获救率也越大
data.pivot_table(index='Parch',values='Survived')
Survived
Parch
0.343658
1 0.550847
2 0.500000
3 0.600000
4 0.000000
5 0.200000
6 0.000000

至此，我们认为重要特征为Pclass，Sex，Age，Embarked，SibSp，Parch

构造一个新的数据表

#至此，我们认为重要特征为Pclass，Sex，Age，Embarked，SibSp，Parch
#构造一个新的数据表
columns = ['Pclass','Sex','Age','SibSp','Parch','Embarked','Survived','Fare']
new_data = data[columns]
new_data.head()
Pclass Sex Age SibSp Parch Embarked Survived Fare
3 male 22.0 1 S 7.2500
1 1 female 38.0 1 C 1 71.2833
2 3 female 26.0 S 1 7.9250
3 1 female 35.0 1 S 1 53.1000
4 3 male 35.0 S 8.0500

查看缺失值

#查看缺失值
new_data.isnull().sum()
           
Pclass        0
Sex           0
Age         177
SibSp         0
Parch         0
Embarked      2
Survived      0
Fare          0
dtype: int64

填充缺失值，年龄填充为中位数，登船地点填充为众数

#填充缺失值，年龄填充为中位数，登船地点填充为众数
new_data['Age'].fillna(new_data['Age'].median(),inplace = True)
print(new_data['Age'].median())
print(new_data['Embarked'].mode())
# #查看数据表描述
new_data.describe()
           
28.0
0    S
dtype: objec
Pclass Age SibSp Parch Survived Fare
count 891.000000 891.000000 891.000000 891.000000 891.000000 891.000000
mean 2.308642 29.361582 0.523008 0.381594 0.383838 32.204208
std 0.836071 13.019697 1.102743 0.806057 0.486592 49.693429
min 1.000000 0.420000 0.000000 0.000000 0.000000 0.000000
25% 2.000000 22.000000 0.000000 0.000000 0.000000 7.910400
50% 3.000000 28.000000 0.000000 0.000000 0.000000 14.454200
75% 3.000000 35.000000 1.000000 0.000000 1.000000 31.000000
max 3.000000 80.000000 8.000000 6.000000 1.000000 512.329200

查看空值

Pclass      0
Sex         0
Age         0
SibSp       0
Parch       0
Embarked    2
Survived    0
Fare        0
dtype: int64

将登船地点填充

#将登船地点填充
new_data["Embarked"] = new_data["Embarked"].fillna('S')
           
Pclass      0
Sex         0
Age         0
SibSp       0
Parch       0
Embarked    0
Survived    0
Fare        0
dtype: int64

将男性和女性的字符值转化为数值男0，女1

#将男性和女性的字符值转化为数值男0，女1
new_data.loc[new_data['Sex']=='male','Sex'] = 0
new_data.loc[new_data['Sex']=='female','Sex'] = 1

将登船地点对应的字符值转化为数值 C：0，Q:1,S:2

#将登船地点对应的字符值转化为数值  C：0，Q:1,S:2
new_data.loc[new_data['Embarked']=='C','Embarked'] = 0
new_data.loc[new_data['Embarked']=='Q','Embarked'] = 1
new_data.loc[new_data['Embarked']=='S','Embarked'] = 2
new_data.head()
Pclass Sex Age SibSp Parch Embarked Survived Fare
3 22.0 1 2 7.2500
1 1 1 38.0 1 1 71.2833
2 3 1 26.0 2 1 7.9250
3 1 1 35.0 1 2 1 53.1000
4 3 35.0 2 8.0500

第二步：初步建模调整参数

使用线性回归 

#开始建模，使用线性回归
from sklearn.linear_model import LinearRegression
#使用交叉验证方法
from sklearn.model_selection import KFold,cross_val_score
kf = KFold(5,random_state=0)
predictors = ["Pclass", "Sex", "Age", "SibSp", "Parch", "Embarked"]
train_target = new_data['Survived']
LR = LinearRegression()
accuracys=[]
for train,test in kf.split(new_data):
    LR.fit(new_data.loc[train,predictors],new_data.loc[train,'Survived'])
    pred = LR.predict(new_data.loc[test,predictors])
    pred[pred >= 0.60] = 1
    pred[pred < 0.60] = 0
    accuracy = len(pred[pred == new_data.loc[test,'Survived']])/len(test)
    accuracys.append(accuracy)
print(np.mean(accuracys))
           
0.8035653756826313

使用逻辑回归 

#开始建模，使用逻辑回归
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import KFold,cross_val_score
kf = KFold(5,random_state=0)
predictors = ["Pclass", "Sex", "Age", "SibSp", "Parch", "Embarked"]
lr = LogisticRegression(C = 0.1,solver='liblinear',penalty='l2')
lr.fit(new_data[predictors],new_data['Survived'])
print(cross_val_score(lr,new_data[predictors],new_data['Survived'],cv = kf).mean())
accuracys = []
for train,test in kf.split(new_data):
    lr.fit(new_data.loc[train,predictors],new_data.loc[train,'Survived'])
    pred = lr.predict_proba(new_data.loc[test,predictors])
#     print(pred.shape)
    new_pred = pred[:,1]
#     print(new_pred)
    new_pred[new_pred >= 0.50] = 1
    new_pred[new_pred < 0.50] = 0
    accuracy = len(new_pred[new_pred == new_data.loc[test,'Survived']])/len(test)
    accuracys.append(accuracy)
print(np.mean(accuracys))
           
0.7956939300734418
0.7956939300734418

使用决策树 

#开始建模，使用决策树
from sklearn import tree
dt = tree.DecisionTreeClassifier(min_samples_split=4, min_samples_leaf=4)
kf = KFold(5,random_state=0)
accuracys = []
for train,test in kf.split(new_data):
    dt.fit(new_data.loc[train,predictors],new_data.loc[train,'Survived'])
    pred = dt.predict(new_data.loc[test,predictors])
    accuracy = len(pred[pred == new_data.loc[test,'Survived']])/len(test)
    accuracys.append(accuracy)
print(np.mean(accuracys))
print(cross_val_score(dt,new_data[predictors],new_data['Survived'],cv=kf).mean())
           
0.804758018956751
0.8036344234511331

第三步：使用集成学习算法，随机森林 

#开始建模，使用随机森林
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import KFold
from sklearn.model_selection import cross_val_score
alg = RandomForestClassifier(random_state=1, n_estimators=80, min_samples_split=4, min_samples_leaf=4)
predictors = ["Pclass", "Sex", "Age", "SibSp", "Parch", "Embarked"]
kf = KFold(5,random_state=0)
scores = cross_val_score(alg, new_data[predictors], new_data["Survived"], cv=kf)  #方法一
print(scores.mean())
accuracys = []   #方法二
for train,test in kf.split(new_data):
    alg.fit(new_data.loc[train,predictors],new_data.loc[train,'Survived'])
    pred = alg.predict(new_data.loc[test,predictors])
    accuracy = len(pred[pred == new_data.loc[test,'Survived']])/len(test)
    accuracys.append(accuracy)
print(np.mean(accuracys))
           
0.820475801895675
0.820475801895675
机器学习 泰坦尼克 线性回归 逻辑回归 决策树 随机森林
上一篇: 接触感知机器人设计的端到端可微框架
下一篇: 压缩感知重构算法之CoSaMP算法python实现

继续阅读