泰坦尼克号乘客生存预测

泰坦尼克号乘客获救预测

1. 案例背景

泰坦尼克号沉船事故是世界上最著名的沉船事故之一。1912年4月15日，在她的处女航期间，泰坦尼克号撞上冰山后沉没，造成2224名乘客和机组人员中超过1502人的死亡。这一轰动的悲剧震惊了国际社会，并导致更好的船舶安全法规。 事故中导致死亡的一个原因是许多船员和乘客没有足够的救生艇。然而在被获救群体中也有一些比较幸运的因素；一些人群在事故中被救的几率高于其他人，比如妇女、儿童和上层阶级。 这个Case里，我们需要分析和判断出什么样的人更容易获救。最重要的是，要利用机器学习来预测出在这场灾难中哪些人会最终获救；

此项目数据集分为2份数据集titanic_train.csv和titanic_test.csv
titanic_train.csv: 训练集，共计891条数据
titanic_test.csv: 测试集，共计418条数据

PassengerId 是数据唯一序号；Survived 是存活情况，为预测标记特征；剩下的10个是原始特征数据。

首先先看看数据

#数据分析库
import pandas as pd
#科学计算库
import numpy as np 
from pandas import Series,DataFrame

data_train = pd.read_csv("datalab/1386/titanic_train.csv")
data_test = pd.read_csv("datalab/1386/titanic_test.csv")

data_train.info()

从数据的信息可以发现，年龄和客舱号数据有丢失。
我们现在对缺失值进行处理，因为年龄的缺失数量比较少，所以直接用中位数进行填充，而客舱号因为缺失值太多，如果进行填充会带来很大的误差，所以对客舱号的不进行处理。

#Age列中的缺失值用Age中位数进行填充
data_train["Age"] = data_train['Age'].fillna(data_train['Age'].median())  
data_train.describe()

#线性回归
from sklearn.linear_model import LinearRegression   

#训练集交叉验证，得到平均值
#from sklearn.cross_validation import KFold 
from sklearn.model_selection import KFold
 
#选取简单的可用输入特征
predictors = ["Pclass","Age","SibSp","Parch","Fare"]     
 
#初始化线性回归算法
alg = LinearRegression()
#样本平均分成3份，3折交叉验证
#我有试着改变交叉验证的折数，发现3折的时候准确率比较高。
kf = KFold(n_splits=3,shuffle=False,random_state=1) 

predictions = []
for train,test in kf.split(data_train):
	# train，test 是从data_train分出来的数据，和n_split有关。
	train_predictors = (data_train[predictors].iloc[train,:])
	train_target = data_train["Survived"].iloc[train]
	# 开始训练数据
	alg.fit(train_predictors,train_target)
	#进行预测
	test_predictions = alg.predict(data_train[predictors].iloc[test,:])
	predictions.append(test_predictions)

import numpy as np
 
#拼接predictions
predictions = np.concatenate(predictions,axis=0)
# 把数据二值化
predictions[predictions>.5] = 1
predictions[predictions<=.5] = 0
accuracy = sum(predictions == data_train["Survived"]) / len(predictions)
print ("准确率为: ", accuracy)

from sklearn.model_selection import cross_val_score
#逻辑回归
from sklearn.linear_model import LogisticRegression   

#初始化逻辑回归算法
LogRegAlg=LogisticRegression(random_state=1)
re = LogRegAlg.fit(data_train[predictors],data_train["Survived"])

#使用sklearn库里面的交叉验证函数获取预测准确率分数
scores =cross_val_score(LogRegAlg,data_train[predictors],data_train["Survived"],cv=3)

#使用交叉验证分数的平均值作为最终的准确率
print("准确率为: ",scores.mean())

增加特征Sex和Embarked列，查看对预测的影响

#Sex性别列处理：male用0，female用1
data_train.loc[data_train["Sex"] == "male","Sex"] = 0
data_train.loc[data_train["Sex"] == "female","Sex"] = 1

#缺失值用最多的S进行填充
data_train["Embarked"] = data_train["Embarked"].fillna('S') 
#地点用0,1,2
data_train.loc[data_train["Embarked"] == "S","Embarked"] = 0    
data_train.loc[data_train["Embarked"] == "C","Embarked"] = 1
data_train.loc[data_train["Embarked"] == "Q","Embarked"] = 2

增加2个特征Sex和Embarked，继续使用逻辑回归算法进行预测

predictors = ["Pclass","Sex","Age","SibSp","Parch","Fare","Embarked"]  

LogRegAlg=LogisticRegression(random_state=1)
#Compute the accuracy score for all the cross validation folds.(much simpler than what we did before!)
re = LogRegAlg.fit(data_train[predictors],data_train["Survived"])
scores = model_selection.cross_val_score(LogRegAlg,data_train[predictors],data_train["Survived"],cv=3)
#Take the mean of the scores (because we have one for each fold)
print("准确率为: ",scores.mean())

通过增加了2个特征，模型的准确率提高到78.78%，说明好的特征有利于提升模型的预测能力。

#新增：对测试集数据进行预处理，并进行结果预测
#Age列中的缺失值用Age均值进行填充
data_test["Age"] = data_test["Age"].fillna(data_test["Age"].median())
#Fare列中的缺失值用Fare最大值进行填充
data_test["Fare"] = data_test["Fare"].fillna(data_test["Fare"].max()) 

#Sex性别列处理：male用0，female用1
data_test.loc[data_test["Sex"] == "male","Sex"] = 0
data_test.loc[data_test["Sex"] == "female","Sex"] = 1
#缺失值用最多的S进行填充
data_test["Embarked"] = data_test["Embarked"].fillna('S') 
#地点用0,1,2
data_test.loc[data_test["Embarked"] == "S","Embarked"] = 0    
data_test.loc[data_test["Embarked"] == "C","Embarked"] = 1
data_test.loc[data_test["Embarked"] == "Q","Embarked"] = 2

test_features = ["Pclass","Sex","Age","SibSp","Parch","Fare","Embarked"] 
#构造测试集的Survived列，
data_test["Survived"] = -1

test_predictors = data_test[test_features]
data_test["Survived"] = LogRegAlg.predict(test_predictors)

7. 使用随机森林算法

from sklearn.model_selection import cross_val_score
from sklearn.ensemble import RandomForestClassifier
 
predictors=["Pclass","Sex","Age","SibSp","Parch","Fare","Embarked"]
 

#10棵决策树，停止的条件：样本个数为2，叶子节点个数为1
alg=RandomForestClassifier(random_state=1,n_estimators=10,min_samples_split=2,min_samples_leaf=1) 

kf=model_selection.KFold(n_splits=3,shuffle=False, random_state=1)

scores=cross_val_score(alg,data_train[predictors],data_train["Survived"],cv=kf)
print(scores)
print(scores.mean())

#30棵决策树，停止的条件：样本个数为2，叶子节点个数为1
alg=RandomForestClassifier(random_state=1,n_estimators=30,min_samples_split=2,min_samples_leaf=1) 

kf=model_selection.KFold(n_splits=10,shuffle=False,random_state=1)

scores=model_selection.cross_val_score(alg,data_train[predictors],data_train["Survived"],cv=kf)
 
print(scores)

print(scores.mean())

PS:此文是参考天池竞赛的入门级文章，加上自己的理解所写，如要找到原文：泰坦尼克号分析