任务8，泰坦尼克号

  1 import pandas as pd
  2 train = pd.read_csv(r‘C:\Users\tanha\Desktop\机器学习\train.csv‘)
  3 test = pd.read_csv(r‘C:\Users\tanha\Desktop\机器学习\test.csv‘)
  4 print(train)
  5 
  6 traindata = train[[‘Pclass‘,‘Sex‘,‘Age‘,‘Embarked‘]].copy()
  7 #trainlabel = pd.Series(train[‘Survived‘])
  8 trainlabel = train[‘Survived‘].copy()#目标值
  9 #为何要.copy()? https://www.jianshu.com/p/72274ccb647a
 10 
 11 #对性别以0-1取代，male-1，female-0
 12 traindata.loc[traindata.Sex ==‘female‘, ‘Sex‘] = 0
 13 traindata.loc[traindata.Sex ==‘male‘, ‘Sex‘] = 1
 14 #对城市也同样用0，1，2取代S,C,Q
 15 traindata.loc[traindata.Embarked ==‘S‘, ‘Embarked‘] = 0
 16 traindata.loc[traindata.Embarked ==‘C‘, ‘Embarked‘] = 1
 17 traindata.loc[traindata.Embarked ==‘Q‘, ‘Embarked‘] = 2
 18 
 19 #查看缺失值情况
 20 traindata.info()
 21 
 22 #显示Age的缺失值
 23 traindata[traindata.Age.isnull()]
 24 #显示Embarked的缺失值
 25 traindata[traindata.Embarked.isnull()]
 26 
 27 #填充Age的缺失值为0时
 28 #traindata.loc[traindata.Age.isnull(), ‘Age‘] = 0
 29 #填充Age的缺失值为该列平均值mean()
 30 #中位数median()
 31 print(traindata.Age.mean())
 32 traindata.loc[traindata.Age.isnull(), ‘Age‘] = traindata.Age.fillna(round(traindata.Age.mean(),1))
 33 #填充Embarked的缺失值为0,s
 34 traindata.loc[traindata.Embarked.isnull(), ‘Embarked‘] = 0
 35 traindata.head(10)
 36 
 37 from sklearn.model_selection import train_test_split
 38 train_data, test_data, train_label, test_label = train_test_split(traindata, trainlabel, random_state=7, test_size=0.3 )
 39 print(len(train_data), len(test_data),len(train_label))
 40 type(train_data)
 41 type(train_label)
 42 
 43 #选出一个K的值
 44 #把训练集分为训练数据和验证数据
 45 import numpy as np
 46 from sklearn.model_selection import KFold
 47 from sklearn.neighbors import KNeighborsClassifier
 48 
 49 X = np.array(traindata[[‘Pclass‘, ‘Sex‘, ‘Age‘,‘Embarked‘]])
 50 #y = np.array(trainlabel[‘Survived‘])
 51 y = trainlabel.values.ravel()
 52 #已经把X,y化为列表
 53 
 54 kf = KFold(n_splits=891,shuffle=True) #如果要使用留一法交叉验证，那n_splits=样本数量
 55 
 56 ks = [1,3,5,7,9,11,13,15,17,19,20]
 57 
 58 best_k = ks[0]
 59 best_score = 0
 60 
 61 for k in ks:
 62     curr_score = 0    
 63     
 64     for train_index, valid_index in kf.split(X):
 65         clf = KNeighborsClassifier(n_neighbors=k)
 66         clf.fit(X[train_index], y[train_index])
 67         curr_score += clf.score(X[valid_index], y[valid_index])
 68     
 69     avg_score = curr_score/891 #前面n_splits=5，5折交叉验证
 70     
 71     if avg_score>best_score:
 72         best_score = avg_score
 73         best_k = k
 74     print("current best score is: %.2f"%best_score, "best k: %d"%best_k)
 75 
 76 print("after cross validtion, the final best k is: %d"%best_k)
 77 
 78 from sklearn.neighbors import KNeighborsClassifier
 79 clf = KNeighborsClassifier(n_neighbors=3)
 80 clf.fit(traindata, trainlabel)
 81 pred = clf.predict(test_data)
 82 from sklearn.metrics import accuracy_score
 83 accuracy_score(pred, test_label)
 84 
 85 testdata = test[[‘Pclass‘,‘Sex‘,‘Age‘,‘Embarked‘]].copy()
 86 #对性别以0-1取代，male-1，female-0
 87 testdata.loc[testdata.Sex ==‘female‘, ‘Sex‘] = 0
 88 testdata.loc[testdata.Sex ==‘male‘, ‘Sex‘] = 1
 89 
 90 testdata.loc[testdata.Embarked ==‘S‘, ‘Embarked‘] = 0
 91 testdata.loc[testdata.Embarked ==‘C‘, ‘Embarked‘] = 1
 92 testdata.loc[testdata.Embarked ==‘Q‘, ‘Embarked‘] = 2
 93 
 94 #查看缺失值情况
 95 testdata.info()
 96 #显示Age的缺失值
 97 testdata[testdata.Age.isnull()]
 98 
 99 #填充Embarked的缺失值为0
100 testdata.loc[testdata.Embarked.isnull(), ‘Embarked‘] = 0
101 
102 #填充Age的缺失值为该列平均值
103 testdata.loc[testdata.Age.isnull(), ‘Age‘] = testdata.Age.fillna(round(testdata.Age.mean(),1))
104 testdata.head(100)
105 
106 import numpy as np
107 result0 = clf.predict(testdata)
108 index = test[‘PassengerId‘]
109 sumission = pd.DataFrame({‘PassengerId‘:test[‘PassengerId‘].values,‘Survived‘:result0.astype(np.int32)})
110 
111 sumission[‘Survived‘] = result0
112 sumission = sumission.reset_index(drop=True)
113 sumission.to_csv(r‘C:\Users\tanha\Desktop\机器学习\submission6.csv‘)