数据分析——泰坦尼克号预测

之前在学校做过课程设计，但是对流程比较一知半解，现在看完了机器学习实战这本书，带着自己的理解重新做一遍。

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

观察数据的具体情况，可以发现年龄变量Age和Cabin有缺失，然后Name，sex，Ticket，cabin和Embark是object类型，在后续的数据处理中要进行调整。

data_train = pd.read_csv(r'C:/Users/train.csv')
data_train.info()

<class 'pandas.core.frame.dataframe'>
RangeIndex: 891 entries, 0 to 890
Data columns (total 12 columns):
 #   Column       Non-Null Count  Dtype
 0   PassengerId  418 non-null    int64
 1   Pclass       418 non-null    int64
 2   Name         418 non-null    object
 3   Sex          418 non-null    object
 4   Age          332 non-null    float64
 5   SibSp        418 non-null    int64
 6   Parch        418 non-null    int64
 7   Ticket       418 non-null    object
 8   Fare         418 non-null    float64
 9   Cabin        91 non-null     object
 10  Embarked     418 non-null    object
dtypes: float64(2), int64(4), object(5)
memory usage: 36.0+ KB
</class>

把索引设置为乘客编号

test_process = test_process.set_index(['PassengerId'])
test_process

现在测试集长这样

PclassNameSexAgeSibSpParchTicketFareEmbarkedCalledName_lengthFirst_namePassengerId8923Kelly, Mr. Jamesmale34003309117.8292QMr16Kelly8933Wilkes, Mrs. James (Ellen Needs)female47103632727.0000SMr32Wilkes8942Myles, Mr. Thomas Francismale62002402769.6875QMr25Myles8953Wirz, Mr. Albertmale27003151548.6625SMr16Wirz8963Hirvonen, Mrs. Alexander (Helga E Lindqvist)female2211310129812.2875SMr44Hirvonen…………………………………13053Spector, Mr. Woolfmale2500A.5. 32368.0500SMr18Spector13061Oliva y Ocana, Dona. Ferminafemale3900PC 17758108.9000CNaN28Oliva y Ocana13073Saether, Mr. Simon Sivertsenmale3800SOTON/O.Q. 31012627.2500SMr28Saether13083Ware, Mr. Frederickmale25003593098.0500SMr19Ware13093Peter, Master. Michael Jmale2211266822.3583CNaN24Peter

418 rows × 12 columns

缺失值处理

本次数据的缺失应该是完全随机的，不依赖于其他完全变量，所以可以采取删除和填补两种方式。cabin缺失过多，直接删除这一特征，不放心的话可以计算一些相关度或者画图看看情况。

train_process = data_train.drop(['Cabin'],axis=1)

from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import GridSearchCV
Age_df = train_process[['Age','Survived','Pclass','SibSp','Parch','Fare']]
UnknowAge = Age_df[Age_df.Age.isnull()].values
KnowAge = Age_df[Age_df.Age.notnull()].values

y_train = KnowAge[:,0]
x_train = KnowAge[:,1:]
rfr = RandomForestRegressor(n_estimators=500,random_state=42)
rfr.fit(x_train,y_train)
predictedAges = rfr.predict(UnknowAge[:,1::])
Age_df.loc[ (Age_df.Age.isnull()), 'Age' ] = predictedAges
train_process.Age=Age_df.Age.astype(int)

年龄缺失值使用随机森林进行填补，建立回归方程进行拟合。

测试集也要删除cabin变量和进行年龄缺失值的填补。

test_process = data_test.drop(['Cabin'],axis=1)
test_process.info()

<class 'pandas.core.frame.dataframe'>
RangeIndex: 418 entries, 0 to 417
Data columns (total 10 columns):
 #   Column       Non-Null Count  Dtype
 0   Pclass       891 non-null    int64
 1   Sex          891 non-null    float64
 2   Age          891 non-null    int32
 3   SibSp        891 non-null    int64
 4   Parch        891 non-null    int64
 5   Ticket       891 non-null    float64
 6   Fare         891 non-null    float64
 7   Embarked     891 non-null    float64
 8   Called       891 non-null    float64
 9   Name_length  891 non-null    float64
 10  First_name   891 non-null    float64
dtypes: float64(7), int32(1), int64(3)
memory usage: 73.2 KB
</class>

投票法

先看看投票法

lr_clf = LogisticRegression(penalty='l1',solver='saga',n_jobs=-1,max_iter=20000)
rnd_clf = RandomForestClassifier(n_estimators=300,max_depth=8,min_samples_leaf=1,min_samples_split=5,random_state=42)
svm_clf = SVC(C=2,kernel='poly',random_state=42,probability=True)
voting_clf = VotingClassifier(estimators=[('lr',lr_clf),('rf',rnd_clf),('scv',svm_clf)],voting='soft')
voting_clf.fit(X_train_encoded,y_train)

  VotingClassifier(estimators=[('lr',
                                  LogisticRegression(max_iter=20000, n_jobs=-1,
                                                     penalty='l1', solver='saga')),
                                 ('rf',
                                  RandomForestClassifier(max_depth=8,
                                                         min_samples_split=5,
                                                         n_estimators=300,
                                                         random_state=42)),
                                 ('scv',
                                  SVC(C=2, kernel='poly', probability=True,
                                      random_state=42))],
                     voting='soft')

y_test = pd.read_csv(r'C:/Users/gender_submission.csv')

y_test = y_test['Survived']

from sklearn.metrics import accuracy_score

for clf in (lr_clf,rnd_clf,svm_clf,voting_clf):
    clf.fit(X_train_encoded,y_train)
    y_pred = clf.predict(X_test_encoded)
    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))

LogisticRegression 0.6961722488038278
RandomForestClassifier 0.80622009569378
SVC 0.6363636363636364
VotingClassifier 0.8110047846889952

再试试XGBoost，果然效果比较好。

XGBoost

import xgboost
from sklearn.metrics import mean_squared_error
xgb_reg = xgboost.XGBRFRegressor(random_state=42)
xgb_reg.fit(X_train_encoded,y_train)
y_pred = xgb_reg.predict(X_test_encoded)
val_error=mean_squared_error(y_test,y_pred)
print("Validation MSE:", val_error)

Validation MSE: 0.5023153196818051

Original: https://blog.csdn.net/weixin_43925467/article/details/124055489
Author: aka.炼金术士
Title: 数据分析——泰坦尼克号预测