虽然理论知识学了很多，但是实际操作还没有积累，现在每天积累一题。
---------------------------不积跬步无以至千里---------------------------------------
Titanic的数据分为test.csv和train.csv，每一行row代表一个乘客的详细信息，每一列column代表一个feature，最后一列是存活信息，1代表存活，0代表没存活。那么，需要用train数据去训练模型，拿到好模型和参数后去测试test，得到test训练之后每个人的存活与否，再与原存活信息比对，看是否一致。
一：导入包
#数据处理和整理 import pandas as pd import numpy as np import random as rnd #画图 import matplotlib.pyplot as plt import seaborn as sns %matplotlib inline #导入的是二分类相关模型，模型融合用的是随机森林。 from sklean.linear_model import LogisticRegression from sklearn.svm import SVC, LinearSVC from sklearn.naive_bayes import GaussianNB from sklearn.linear_model import Perceptron from sklearn.linear_model import SGDClassifier from sklearn.ensemble import RandomForestClassifier from sklearn.neighbors import KNeighborsClassifier from sklearn.tree import DecisionTreeClassifier
二：获得数据
train_df = pd.read_csv('train.csv')
test_df = pd.read_csv('test.csv')
combine = [train_df,test_df] #list中可以放入str，dict,int,float,也可以放入dataframe。可以用同一种处理迭代在两个数据上，保持数据一致性。
三：分析数据
查看每一列的特征名称
print(train_df.columns.values) #也可以list[dataframe.columns.values], list[dataframe]
['PassengerId' 'Survived' 'Pclass' 'Name' 'Sex' 'Age' 'SibSp' 'Parch' 'Ticket' 'Fare' 'Cabin' 'Embarked']
看一下每一列所包含的数据是什么？数值型数据还是分类型数据？数值型数据就可以用来计算，分类型数据可以用来定性。
train_df.head() #调用方法，产生数值需要括号，谨记类的调用知识。如果看头十行，head(10)
那么，对数据要上一下几点的心：
- Which features are available in the dataset?
- Which features are categorical?
- Which features are numerical?
- Which features are mixed data types?
- Which features may contain errors or typos?
- Which features contain blank, null or empty values?
- What are the data types for various features?
train_df.info
查看每一列和survived的联系
思想：过滤式特征选择 统计特征对于结果的贡献程度。
简单一点的方法：groupby
train[['Pclass','Survived']].groupby(['Pclass',as_index=False],mean().sort_values(['Survivued'],ascending = False)
把'Pclass'特征和结果'Survived'组合起来，通过Pclass作为标签，计算年龄的存活率。这个有一点不好，就是Pclass有三个值，1,2,3，这三个值如果作为x点乘θ的贡献其实是平等的，但是1,2,3却赋予了它们不等的值，需要做one-hot处理。
查看各列的组合和survived的联系
grid = sns.FacetGrid(train_df, row='Embarked', col='Survived', size=2.2, aspect=1.6) grid.map(sns.barplot, 'Sex', 'Fare', alpha=.5, ci=None) grid.add_legend()
四：数据清洗
去掉脏数据和对结果没贡献的数据。
train_df = train_df.drop(['Ticket','Cabin'],axis=1) #axis=1去掉columns test_df = test_df.drop(['Ticket','Cabin'],axis = 1) combine = [train_df,test_df] #去掉数据需要单独做，再combine。
生成一个新特征，新特征提取自旧特征，相当于是旧特征的再加工。
We want to analyze if Name feature can be engineered to extract titles and test correlation between titles and survival, before dropping Name and PassengerId features. 这个有点玄学，但是能学到一点，如何通过dataframe做re的事情。外国人的名字是Jack.Johson。那么，通过匹配前面Jack.就可以拿到他们的title。
for dataset in combine:
dataset['Title'] = dataset.Name.str.extract('([A-za-z]+)\.',expand=False) #dataset.ColumnName也可以用dataset['ColumnName']
pd.crosstab(train_df['Title'],train_df['Sex']) #第一个参数是指定index，第二个参数是指定column
Sex female male Title Capt 0 1 Col 0 2 Countess 1 0 Don 0 1 Dr 1 6 Jonkheer 0 1 Lady 1 0 Major 0 2 Master 0 40还可以这样更进一步玩：
df = pd.crosstab(df['Title'], df['Sex'],values=df['Survived'],aggfunc=sum) #把对应项的值求和
把特征的值替换掉：
for dataset in combine:
dataset['Title'] = dataset['Title'].replace(['Lady', 'Countess','Capt', 'Col',\
'Don', 'Dr', 'Major', 'Rev', 'Sir', 'Jonkheer', 'Dona'], 'Rare') #replace(x,y) 把x替换成y
dataset['Title'] = dataset['Title'].replace('Mlle', 'Miss')
dataset['Title'] = dataset['Title'].replace('Ms', 'Miss')
dataset['Title'] = dataset['Title'].replace('Mme', 'Mrs')
train_df[['Title', 'Survived']].groupby(['Title'], as_index=False).mean()
title_mapping = {"Mr": 1, "Miss": 2, "Mrs": 3, "Master": 4, "Rare": 5}
for dataset in combine:
dataset['Title'] = dataset['Title'].map(title_mapping)
dataset['Title'] = dataset['Title'].fillna(0)
train_df.head()
###########################更进一步处理数据##############################
哑变量处理（这是跑完下面的建模分析做对比之后，我第二次再加工的数据，下面的预测都是未做哑变量处理的）
dict_Pclass = {1:'1st',2:'2nd',3:'3rd'}
#总觉得Pclass那边做astype(int)不太好，因为普通乘客3，会员2，尊贵黄金会员1是等价的所以改回来，用str代替了数字1,2,3，并用get_dummies处理。
train_df ["Pclass"] = train_df["Pclass"].map(dict_Pclass)
test_df["Pclass"] = test_df["Pclass"].map(dict_Pclass)
test_df = test_df.drop(['Age*Pclass'],axis = 1)
train_df = pd.get_dummies(train_df) #把dataframe中的str数据分类，形成新特征，包含该新特征的样本是1，不包含是0. test_df = pd.get_dummies(test_df) #好处是更加容易做计算，坏处是对特征稀释。
处理过后对训练集的预测结果：
LR:80.920000000000002 < 81.37 未处理SVC:83.28 < 83.95 未处理
kNN:84.510000000000005 < 84.95
naive_bayes:75.530000000000001 > 73.73perceptron:81.140000000000001 > 74.06linear_SVC:81.930000000000007 > 81.14Decision_Tree:86.760000000000005 = 86.76
Random_Forest:86.640000000000001 = 86.64
讲道理，进行了哑变量变换之后，应该线性模型都会更加精确，不知道为什么LR和SVC，kNN变小了，是不是因为样本数很小，体现不出来？
################总之，初步的处理了一下数据###################
X_train = train_df.drop(['Survived'],axis = 1)
y_train = train_df['Survived']
X_test = test_df.drop(['PassengerId'],axis=1).copy() #传值
#Logistic Regression
logist = LogisticRegression() #初始化类
logist.fit(X_train,y_train) #调用类中的fit函数，训练数据，不需要返回值
y_predict = logist.predict(X_test)
acc_log = round(logist.score(X_train, y_train) * 100, 2)
acc_log
#81.370000000000005
coeff_df = pd.DataFrame(train_df.columns.delete(0))
coeff_df["Correlation"] = pd.Series(logist.coef_[0])
coeff_df.sort_values(by='Correlation', ascending=False)
svc = SVC()
svc.fit(X_train,y_train)
svc_predict = svc.predict(X_test)
acc_svc = round(svc.score(X_train,y_train ) *100,2) #给出该模型的平均正确率
print(acc_svc)
#83.95
kn = KNeighborsClassifier()
kn.fit(X_train,y_train)
y_test = kn.predict(X_test)
acc_knn = round(kn.score(X_train,y_train) * 100, 2)
acc_knn
#84.959999999999994
gaussian = GaussianNB()
gaussian.fit(X_train, y_train)
y_test = gaussian.predict(X_test)
acc_gaus = round(gaussian.score(X_train,y_train) * 100, 2)
acc_gaus
#73.739999999999995
percep = Perceptron()
percep.fit(X_train, y_train)
y_test = percep.predict(X_test)
acc_per = round(percep.score(X_train, y_train) * 100, 2)
acc_per
#74.069999999999993
l_svc = LinearSVC()
l_svc.fit(X_train,y_train)
y_test = l_svc.predict(X_test)
acc_lsvc = round(l_svc.score(X_train, y_train) * 100 , 2)
acc_lsvc
#81.140000000000001
deci_tre = DecisionTreeClassifier()
deci_tre.fit(X_train,y_train)
deci_tre_y_test = deci_tre.predict(X_test)
acc_deci = round(deci_tre.score(X_train, y_train) * 100 , 2)
acc_deci
#86.760000000000005
ran_fo = RandomForestClassifier()
ran_fo.fit(X_train, y_train )
y_test = ran_fo.predict(X_test)
acc_ran_fo = round(ran_fo.score(X_train, y_train) * 100, 2)
acc_ran_fo
#86.640000000000001
submission = pd.DataFrame({
"PassengerId": test_df["PassengerId"],
"Survived":y_test
})
submission.to_csv("Titanic_submission.csv",index=False)