机器学习 交叉验证与网格搜索调参

网格搜索一般是针对参数进行寻优，交叉验证是为了验证训练模型拟合程度。sklearn中的相关API如下：

（1）交叉验证的首要工作：切分数据集train/validation/test

A.)没指定数据切分方式，直接选用cross_val_score按默认切分方式进行交叉验证评估得分，如下图

from sklearn.model_selection import cross_val_score
from sklearn.datasets import load_iris
from sklearn.linear_model import LogisticRegression

iris = load_iris()
logreg = LogisticRegression()

scores = cross_val_score(logreg, iris.data, iris.target, cv=5)
#默认cv=3,没指定默认在训练集和测试集上进行交叉验证
scores

#Output:
#array([ 1.        ,  0.96666667,  0.93333333,  0.9       ,  1.        ])
   
   
1
2
3
4
5
6
7
8
9
10
11
12
13

B.)K折交叉验证KFold

from sklearn.datasets import load_iris 
from sklearn.model_selection import KFold
iris = load_iris()
kfold = KFold(n_splits=5)
cross_val_score(logreg, iris.data, iris.target, cv=kfold)

#Output:
#array([ 1.        ,  0.93333333,  0.43333333,  0.96666667,  0.43333333])
   
   
1
2
3
4
5
6
7
8

from sklearn.datasets import load_iris
from sklearn.model_selection import KFold
kfold = KFold(n_splits=3, shuffle=True, random_state=0)
#shuffle添加了随机扰动,打乱样本顺序,再进行k折切分样本
cross_val_score(logreg, iris.data, iris.target, cv=kfold)

#Output:
#array([ 0.9 ,  0.96,  0.96])
   
   
1
2
3
4
5
6
7
8

C.)留一交叉验证LeaveOneOut(工业实践很少用)

from sklearn.datasets import load_iris
from sklearn.model_selection import LeaveOneOut
loo = LeaveOneOut()
scores = cross_val_score(logreg, iris.data, iris.target, cv=loo)
#cv参数Determines the cross-validation splitting strategy,An object to be used as a cross-validation generator.
print("number of cv iterations: ", len(scores))
print("mean accuracy: ", scores.mean())

#Output:
#number of cv iterations:  150
#mean accuracy:  0.953333333333
   
   
1
2
3
4
5
6
7
8
9
10
11

D.)乱序分割交叉验证ShuffleSplit

from sklearn.datasets import load_iris
from sklearn.model_selection import ShuffleSplit   
#random splits do not guarantee that all folds will be different, although this is still very likely for sizeable datasets.
shuffle_split = ShuffleSplit(test_size=.5, train_size=.5, n_splits=10,random_state=0)
cross_val_score(logreg, iris.data, iris.target, cv=shuffle_split)

#Output:
#array([ 0.84      ,  0.93333333,  0.90666667,  1.        ,  0.90666667,
#        0.93333333,  0.94666667,  1.        ,  0.90666667,  0.88      ])
   
   
1
2
3
4
5
6
7
8
9

E.)数据与分组交叉验证GroupKFold

The same group will not appear in two different folds (the number of distinct groups has to be at least equal to the number of folds).

from sklearn.model_selection import GroupKFold
X = np.array([[1, 2], [3, 4], [5, 6], [7, 8]])
y = np.array([1, 2, 3, 4])
groups = np.array([0, 0, 2, 2])
gkf=GroupKFold(n_splits=2)
for train_index, test_index in gkf.split(X, y, groups):
        print("TRAIN:", train_index, "TEST:", test_index)
        X_train, X_test = X[train_index], X[test_index]
        y_train, y_test = y[train_index], y[test_index]
        print('X_train:',X_train)
        print('X_test:',X_test)
        print('y_train:',y_train)
        print('y_test:',y_test)

Output:
TRAIN: [0 1] TEST: [2 3]
X_train: [[1 2]
 [3 4]]
X_test: [[5 6]
 [7 8]]
y_train: [1 2]
y_test: [3 4]

TRAIN: [2 3] TEST: [0 1]
X_train: [[5 6]
 [7 8]]
X_test: [[1 2]
[3 4]]
y_train: [3 4]
y_test: [1 2]
   
   
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

F.)按样本的标签分层切分StratifiedKFold

（2.）有关模型的参数调优过程，即网格搜索/交叉验证

a.)最简单的网格搜索：两层for循环

# naive grid search implementation
from sklearn.datasets import load_iris
from sklearn.svm import SVC
from sklearn.model_selection import train_test_split
iris = load_iris()
X_train, X_test, y_train, y_test = train_test_split(iris.data, iris.target, random_state=0)
print("Size of training set: %d   size of test set: %d" % (X_train.shape[0], X_test.shape[0]))

best_score = 0
for gamma in [0.001, 0.01, 0.1, 1, 10, 100]:
    for C in [0.001, 0.01, 0.1, 1, 10, 100]:
        # for each combination of parameters
        # train an SVC
        svm = SVC(gamma=gamma, C=C)
        svm.fit(X_train, y_train)
        # evaluate the SVC on the test set 
        score = svm.score(X_test, y_test)
        # if we got a better score, store the score and parameters
        if score > best_score:
            best_score = score
            best_parameters = {'C': C, 'gamma': gamma}

print("best score: ", best_score)
print("best parameters: ", best_parameters)

#Output:
#Size of training set: 112   size of test set: 38
#best score:  0.973684210526
#best parameters:  {'gamma': 0.001, 'C': 100}
   
   
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

在训练集上再切出来一部分作为验证集，用于评估模型，防止过拟合

from sklearn.datasets import load_iris
from sklearn.svm import SVC
from sklearn.model_selection import train_test_split
iris = load_iris()
X_train, X_test, y_train, y_test = train_test_split(iris.data, iris.target, random_state=0)
print("Size of training set: %d   size of test set: %d" % (X_train.shape[0], X_test.shape[0]))

best_score = 0
for gamma in [0.001, 0.01, 0.1, 1, 10, 100]:
    for C in [0.001, 0.01, 0.1, 1, 10, 100]:
        # for each combination of parameters
        # train an SVC
        svm = SVC(gamma=gamma, C=C)
        svm.fit(X_train, y_train)
        # evaluate the SVC on the test set 
        score = svm.score(X_test, y_test)
        # if we got a better score, store the score and parameters
        if score > best_score:
            best_score = score
            best_parameters = {'C': C, 'gamma': gamma}

print("best score: ", best_score)
print("best parameters: ", best_parameters)

#Output:
#Size of training set: 112   size of test set: 38
#best score:  0.973684210526
#best parameters:  {'gamma': 0.001, 'C': 100}
   
   
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

b.)网格搜索内部嵌套了交叉验证

import numpy as np
from sklearn.datasets import load_iris
from sklearn.svm import SVC
from sklearn.model_selection import cross_val_score
iris = load_iris()
X_trainval, X_test, y_trainval, y_test = train_test_split(iris.data, iris.target, random_state=0)   #总集——>训练验证集+测试集
X_train, X_valid, y_train, y_valid = train_test_split(X_trainval, y_trainval, random_state=1)   #训练验证集——>训练集+验证集
best_score = 0
for gamma in [0.001, 0.01, 0.1, 1, 10, 100]:
    for C in [0.001, 0.01, 0.1, 1, 10, 100]:
        svm = SVC(gamma=gamma, C=C)
        scores = cross_val_score(svm, X_trainval, y_trainval, cv=5)   #在训练集和验证集上进行交叉验证
        score = np.mean(scores)   # compute mean cross-validation accuracy
        if score > best_score:   
            best_score = score
            best_parameters = {'C': C, 'gamma': gamma}

# rebuild a model on the combined training and validation set
print('网格搜索for循环<有cross_val_score交叉验证>获得的最好参数组合:',best_parameters)
print(' ')
svmf = SVC(**best_parameters)
svmf.fit(X_trainval, y_trainval)
print('网格搜索<有交叉验证>获得的最好估计器,在训练验证集上没做交叉验证的得分:',svmf.score(X_trainval,y_trainval))#####
print(' ')
scores = cross_val_score(svmf, X_trainval, y_trainval, cv=5)   #在训练集和验证集上进行交叉验证   
print('网格搜索<有交叉验证>获得的最好估计器,在训练验证集上做交叉验证的平均得分:',np.mean(scores))   #交叉验证的平均accuracy   
print(' ')
print('网格搜索<有交叉验证>获得的最好估计器,在测试集上的得分:',svmf.score(X_test,y_test))#####
# print(' ')
# print(' ')
# scoreall = cross_val_score(svmf, iris.data, iris.target, cv=5)  
# print(scoreall ,np.mean(scoreall)) 
   
   
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32

Output:

网格搜索for循环<有cross_val_score交叉验证>获得的最好参数组合: {'gamma': 0.01, 'C': 100}

网格搜索<有交叉验证>获得的最好估计器,在训练验证集上没做交叉验证的得分: 0.982142857143

网格搜索<有交叉验证>获得的最好估计器,在训练验证集上做交叉验证的平均得分: 0.972689629211

网格搜索<有交叉验证>获得的最好估计器,在测试集上的得分: 0.973684210526
   
   
1
2
3
4
5
6
7

c.)构造参数字典,代替双层for循环进行网格搜索

param_grid = {'C': [0.001, 0.01, 0.1, 1, 10, 100],'gamma': [0.001, 0.01, 0.1, 1, 10, 100]}
from sklearn.model_selection import GridSearchCV
from sklearn.svm import SVC
X_trainvalid, X_test, y_trainvalid, y_test = train_test_split(iris.data, iris.target, random_state=0)   #default=0.25
grid_search = GridSearchCV(SVC(), param_grid, cv=5)   #网格搜索+交叉验证
grid_search.fit(X_trainvalid, y_trainvalid)
print('GridSearchCV交叉验证网格搜索字典获得的最好参数组合',grid_search.best_params_)
print(' ')
print('GridSearchCV交叉验证网格搜索获得的最好估计器,在训练验证集上没做交叉验证的得分',grid_search.score(X_trainvalid,y_trainvalid))#####
print(' ')
print('GridSearchCV交叉验证网格搜索获得的最好估计器,在**集上做交叉验证的平均得分',grid_search.best_score_)#?????   
# print(' ')
# print('BEST_ESTIMATOR:',grid_search.best_estimator_)   #对应分数最高的估计器
print(' ')
print('GridSearchCV交叉验证网格搜索获得的最好估计器,在测试集上的得分',grid_search.score(X_test, y_test))#####
   
   
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

Output:

GridSearchCV交叉验证网格搜索字典获得的最好参数组合 {'gamma': 0.01, 'C': 100}

GridSearchCV交叉验证网格搜索获得的最好估计器,在训练验证集上没做交叉验证的得分 0.982142857143

GridSearchCV交叉验证网格搜索获得的最好估计器,在**集上做交叉验证的平均得分 0.973214285714

GridSearchCV交叉验证网格搜索获得的最好估计器,在测试集上的得分 0.973684210526
   
   
1
2
3
4
5
6
7

d.)嵌套交叉验证：字典参数+cross_val_score

from sklearn.datasets import load_iris
from sklearn.model_selection import GridSearchCV
from sklearn.model_selection import cross_val_score
from sklearn.svm import SVC
param_grid = {'C': [0.001, 0.01, 0.1, 1, 10, 100],'gamma': [0.001, 0.01, 0.1, 1, 10, 100]}
scores = cross_val_score(GridSearchCV(SVC(), param_grid, cv=5), iris.data, iris.target, cv=5)   
#选定网格搜索的每一组超参数,对训练集与测试集的交叉验证(cross_val_score没指定数据集合分割的默认情况)
print("Cross-validation scores: ", scores)
print("Mean cross-validation score: ", scores.mean())

#Output:
#Cross-validation scores:  [ 0.96666667  1.          0.96666667  0.96666667  1.        ]
#Mean cross-validation score:  0.98
   
   
1
2
3
4
5
6
7
8
9
10
11
12
13

def nested_cv(X, y, inner_cv, outer_cv, Classifier, parameter_grid):
    outer_scores = []
    # for each split of the data in the outer cross-validation
    # (split method returns indices)
    for training_samples, test_samples in outer_cv.split(X, y):
        # find best parameter using inner cross-validation:网格搜索外层cv
        best_parms = {}
        best_score = -np.inf
        # iterate over parameters
        for parameters in parameter_grid:
            # accumulate score over inner splits
            cv_scores = []
            # iterate over inner cross-validation
            for inner_train, inner_test in inner_cv.split(X[training_samples], y[training_samples]):
                # build classifier given parameters and training data交叉验证内层cv
                clf = Classifier(**parameters)
                clf.fit(X[inner_train], y[inner_train])
                # evaluate on inner test set
                score = clf.score(X[inner_test], y[inner_test])
                cv_scores.append(score)
            # compute mean score over inner folds
            mean_score = np.mean(cv_scores)
            if mean_score > best_score:
                # if better than so far, remember parameters
                best_score = mean_score
                best_params = parameters
        # build classifier on best parameters using outer training set
        clf = Classifier(**best_params)
        clf.fit(X[training_samples], y[training_samples])
        # evaluate 
        outer_scores.append(clf.score(X[test_samples], y[test_samples]))
    return outer_scores

from sklearn.datasets import load_iris
from sklearn.svm import SVC
from sklearn.model_selection import ParameterGrid, StratifiedKFold
param_grid = {'C': [0.001, 0.01, 0.1, 1, 10, 100],'gamma': [0.001, 0.01, 0.1, 1, 10, 100]}
#http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.ParameterGrid.html#sklearn.model_selection.ParameterGrid
#http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.StratifiedKFold.html#sklearn.model_selection.StratifiedKFold
#ParameterGrid是按给定参数字典分配训练集与测试集,StratifiedKFold是分层分配训练集与测试集
nested_cv(iris.data, iris.target, StratifiedKFold(5), StratifiedKFold(5), SVC, ParameterGrid(param_grid))

#Output:
#[0.96666666666666667, 1.0, 0.96666666666666667, 0.96666666666666667, 1.0]