预测Titanic号上的乘客生存概率_03_优化训练集

Posted 2022-07-21 SongpingWang

文章目录

• ​​一、导入数据 数据清洗​​
• ​​二、构建 onehot 编码​​
• ​​三、筛选处理好的变量、拆分数据集​​
• ​​四、训练—预测—筛选最优分类器​​

import numpy as np
import pandas as pd
from sklearn import metrics
from sklearn import model_selection
from sklearn.linear_model import LogisticRegression
from sklearn.svm import SVC,LinearSVC
from sklearn.neighbors import KNeighborsClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
from matplotlib import pyplot as plt
%matplotlib inline
import

一、导入数据 数据清洗

train_src=pd.read_csv(train.csv)
train_src[Data_type]=train
pred_src=pd.read_csv(test.csv)
pred_src[Data_type]=pred
combine=pd.concat([train_src,pred_src]) 

combine[Age]=combine[Age].fillna(train_src[Age].mean())    #对年龄缺失值进行均值填充
combine[Fare]=combine[Fare].fillna(train_src[Fare].mean()) #Fare船票价格空值,平均值替换 

#从姓名中提取头衔
def get_title(name):
    title_search=re.search(, ([A-Za-z]+)\\.,name)
    if title_search:
        return title_search.group(1)
    return np.nan
combine[Title]=combine[Name].apply(get_title)
combine[Title].value_counts(dropna=False) 

combine[Title]=combine[Title].replace([Lady,Capt,Col,Don,Dr,Major,
                                           Rev,Sir,Jonkheer,Dona],Rare)
combine[Title]=combine[Title].replace(Mlle,Miss)
combine[Title]=combine[Title].replace(Ms,Miss)
combine[Title]=combine[Title].replace(Mme,Mrs)
combine[Title].value_counts(dropna=False) 

#兄弟姐妹数，配偶数，父母数，子女数加起来
combine[FamilySize]=combine[SibSp]+combine[Parch]+1
combine[FamilySize].value_counts() 

#从性别、年龄、子女数来判断乘客是否是母亲
combine[is_mother]=0
combine.loc[(combine[Sex]==female) & (combine[Parch]>0) & (combine[Age]>20),is_mother]=1
combine[is_mother].value_counts()

二、构建 onehot 编码

sex_onehot=pd.get_dummies(combine[Sex],drop_first=False,prefix=onehot)
sex_onehot.head() 

combine[Age_group]=np.nan
combine.loc[combine[Age]<=16,Age_group]=Age_0_16
combine.loc[(combine[Age]>16)&(combine[Age]<=32),Age_group]=Age_16_32
combine.loc[(combine[Age]>32)&(combine[Age]<=48),Age_group]=Age_32_48
combine.loc[(combine[Age]>48)&(combine[Age]<=64),Age_group]=Age_48_64
combine.loc[combine[Age]>64,Age_group]=Age_64_
age_group_onehot=pd.get_dummies(combine[Age_group],drop_first=False,prefix=onehot)
age_group_onehot.head() 

title_onehot=pd.get_dummies(combine[Title],drop_first=False,prefix=onehot)
title_onehot.head()

pclass_onehot=pd.get_dummies(combine[Pclass],drop_first=False,prefix=onehot)
pclass_onehot.head()

三、筛选处理好的变量、拆分数据集

combine=pd.concat([combine,sex_onehot,age_group_onehot,title_onehot,pclass_onehot],axis=1)
combine.head() 

pred_X=combine.ix[combine[Data_type]==pred,
                  [SibSp,Parch,Fare,
                   FamilySize,is_mother,onehot_male,onehot_female,
                   onehot_Age_0_16,onehot_Age_16_32,onehot_Age_32_48,
                   onehot_Age_48_64,onehot_Age_64_,
                   onehot_Master,onehot_Miss,onehot_Mr,onehot_Mrs,onehot_Rare,
                   onehot_1,onehot_2,onehot_3]]
pred_id=combine.ix[combine[Data_type]==pred,PassengerId]

X_train,X_test,y_train,y_test = model_selection\\
                                .train_test_split(train_X,train_y, 
                                 test_size=0.3,random_state=42)

四、训练—预测—筛选最优分类器

def classifier_func(X_train,y_train,X_test,y_test):
    #LR
    lr=LogisticRegression()
    lr.fit(X_train,y_train)
    lr_training = lr.score(X_train,y_train)
    lr_testing  = lr.score(X_test,y_test) 
    lr_error_matrix = abs(lr_training-lr_testing)  #lr残差
    print("LR   ：训练准确率:，预测准确率:，残差:".format(lr_training,lr_testing,lr_error_matrix))

    #SVM
    svc=SVC()
    svc.fit(X_train,y_train)
    SVM_training = svc.score(X_train,y_train) 
    SVM_testing = svc.score(X_test,y_test) 
    SVM_error_matrix = abs(SVM_training-SVM_testing) 
    print("SVM  ：训练准确率:，预测准确率:，残差:".format(SVM_training,SVM_testing,SVM_error_matrix))

    #KNN
    knn=KNeighborsClassifier(n_neighbors=3)
    knn.fit(X_train,y_train)
    KNN_training = knn.score(X_train,y_train)
    KNN_testing = knn.score(X_test,y_test) 
    KNN_error_matrix = abs(KNN_training-KNN_testing)
    print("KNN  ：训练准确率:， 预测准确率:，残差:".format(KNN_training,KNN_testing,KNN_error_matrix))

    #决策树
    dtree=DecisionTreeClassifier()
    dtree.fit(X_train,y_train)
    DTree_training = dtree.score(X_train,y_train)
    DTree_testing = dtree.score(X_test,y_test) 
    DTree_error_matrix = abs(DTree_training-DTree_testing)
    print("DTree：训练准确率:，预测准确率:，残差:".format(DTree_training,DTree_testing,DTree_error_matrix))

    #随机森林
    random_forest=RandomForestClassifier(n_estimators=10)
    random_forest.fit(X_train,y_train)
    RF_training = random_forest.score(X_train,y_train) 
    RF_testing = random_forest.score(X_test,y_test)
    RF_error_matrix = abs(RF_training-RF_testing)
    print("RF   ：训练准确率:，预测准确率:，残差:".format(RF_training,RF_testing,RF_error_matrix))

    return 

classifier_func(X_train,y_train,X_test,y_test)  
-------------------------------------------
输出：
LR   ：训练准确率:0.8314606741573034，预测准确率:0.8171641791044776，残差:0.01429649505282582
SVM  ：训练准确率:0.8346709470304976，预测准确率:0.8059701492537313，残差:0.028700797776766263
KNN  ：训练准确率:0.8587479935794543，预测准确率:0.7723880597014925，残差:0.08635993387796181
DTree：训练准确率:0.9502407704654896，预测准确率:0.8097014925373134，残差:0.14053927792817622
RF   ：训练准确率:0.9325842696629213，预测准确率:0.7947761194029851，残差:0.1378081502599362 
--------------------------------------------
# 准备预测数据
pred_X=combine.ix[combine[Data_type]==pred,
                  [SibSp,Parch,Fare,
                   FamilySize,is_mother,onehot_male,onehot_female,
                   onehot_Age_0_16,onehot_Age_16_32,onehot_Age_32_48,onehot_Age_48_64,onehot_Age_64_,
                   onehot_Master,onehot_Miss,onehot_Mr,onehot_Mrs,onehot_Rare,
                   onehot_1,onehot_2,onehot_3]]
pred_id=combine.ix[combine[Data_type]==pred,PassengerId] 

pred_y_pre=lr.predict(pred_X)
test_pred=pd.DataFrame(PassengerId:pred_id,
                       Survived:pred_y_pre.astype(int))
test_pred.to_csv(test_pred_2.csv,index=False)

from sklearn.ensemble import VotingClassifier  #多模型投票融合器
votingcf=VotingClassifier(estimators=[(lr,lr),(knn,knn),(decisiontree,dtree),
                                     (rf,random_forest),(svm,svc)],voting=hard) 
                                     
#训练集投票后的结果
votingcf.fit(X_train,y_train)
votingcf.score(X_train,y_train)  

#测试集投票后的结果
votingcf.fit(X_test,y_test)
votingcf.score(X_test,y_test) 

pred_y_pre=votingcf.predict(pred_X)
submit=pd.DataFrame(PassengerId:pred_id,
                    Survived:pred_y_pre.astype(int))
submit.to_csv(test_pred_3.csv,index=False)