ML - 集成学习 Code

---

集成学习

import numpy as np
import matplotlib.pyplot as plt
 
from sklearn import datasets

X, y = datasets.make_moons(n_samples=500, noise=0.3, random_state=42)  # n_samples 不赋值的话，默认生成 100个数据。
  
from sklearn.model_selection import train_test_split 
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)
 
plt.scatter(X[y==0,0], X[y==0,1])
plt.scatter(X[y==1,0], X[y==1,1])
plt.show()

---

手动实现集成学习

from sklearn.linear_model import LogisticRegression

log_clf = LogisticRegression()
log_clf.fit(X_train, y_train)
log_clf.score(X_test, y_test)  # 0.86399999999999999

from sklearn.svm import SVC

svm_clf = SVC()
svm_clf.fit(X_train, y_train)
svm_clf.score(X_test, y_test)  # 0.88800000000000001

from sklearn.tree import DecisionTreeClassifier

dt_clf = DecisionTreeClassifier(random_state=666)
dt_clf.fit(X_train, y_train)
dt_clf.score(X_test, y_test)  # 0.86399999999999999

y_predict1 = log_clf.predict(X_test)
y_predict2 = svm_clf.predict(X_test)
y_predict3 = dt_clf.predict(X_test)

# 少数服从多数；至少有两个模型认为它是 1，就判断为 1，否则为 0.
y_predict = np.array((y_predict1 + y_predict2 + y_predict3) >= 2, dtype='int')  
 
y_predict[:10]
# array([1, 0, 0, 1, 1, 1, 0, 0, 0, 0])
 
from sklearn.metrics import accuracy_score

accuracy_score(y_test, y_predict)  # 0.89600000000000002

---

Voting

使用 Hard VotingClassifier

from sklearn.ensemble import VotingClassifier # ensemble 就是集成学习的意思 

voting_clf = VotingClassifier(estimators=[
    ('log_clf', LogisticRegression()), 
    ('svm_clf', SVC()),
    ('dt_clf', DecisionTreeClassifier(random_state=666))],
                             voting='hard')
 
voting_clf.fit(X_train, y_train)
voting_clf.score(X_test, y_test)  # 0.89600000000000002

VotingClassifier : https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.VotingClassifier.html

---

使用 Soft Voting Classifier

voting_clf2 = VotingClassifier(estimators=[
    ('log_clf', LogisticRegression()), 
    ('svm_clf', SVC(probability=True)),  # SVC 默认不计算概率；需要把参数 probability 设置为 true 才行。
    ('dt_clf', DecisionTreeClassifier(random_state=666))], 
                             voting='soft')
 
voting_clf2.fit(X_train, y_train)
voting_clf2.score(X_test, y_test)  # 0.91200000000000003

---

Bagging 和 Pasting

from sklearn.tree import DecisionTreeClassifier # 这里使用决策树模型，因为更能产生差异比较大的子模型；所以要在集成学习中，集成成百上千个子模型，首选决策树  
from sklearn.ensemble import BaggingClassifier

# n_estimators：集成多少个子模型；max_samples：每个子模型看多少个样本数据；bootstrap：是否放回取样，true:放回。  Bagging 和 Pasting 在sklearn 中统一使用 BaggingClassifier，仅用 bootstrap 区分。
bagging_clf = BaggingClassifier(DecisionTreeClassifier(),
                           n_estimators=500, max_samples=100,
                           bootstrap=True)
bagging_clf.fit(X_train, y_train)
bagging_clf.score(X_test, y_test)  # 0.91200000000000003

---

# 集成更多子模型，运行会稍微慢一些；理论上子模型越大，准确率越高
bagging_clf = BaggingClassifier(DecisionTreeClassifier(),
                           n_estimators=5000, max_samples=100,
                           bootstrap=True)
bagging_clf.fit(X_train, y_train)
bagging_clf.score(X_test, y_test)  # 0.92000000000000004

---

oob

from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import BaggingClassifier

bagging_clf = BaggingClassifier(DecisionTreeClassifier(),
                               n_estimators=500, max_samples=100,
                               bootstrap=True, oob_score=True) # 构造的使用需要增加 oob_score，告诉这个类需要记录oob
bagging_clf.fit(X, y)

BaggingClassifier(base_estimator=DecisionTreeClassifier(class_weight=None, criterion='gini', max_depth=None,
            max_features=None, max_leaf_nodes=None,
            min_impurity_decrease=0.0, min_impurity_split=None,
            min_samples_leaf=1, min_samples_split=2,
            min_weight_fraction_leaf=0.0, presort=False, random_state=None,
            splitter='best'),
         bootstrap=True, bootstrap_features=False, max_features=1.0,
         max_samples=100, n_estimators=500, n_jobs=1, oob_score=True,
         random_state=None, verbose=0, warm_start=False)

bagging_clf.oob_score_  # 0.91800000000000004

---

并行化处理（n_jobs）

%%time
bagging_clf = BaggingClassifier(DecisionTreeClassifier(),
                               n_estimators=500, max_samples=100,
                               bootstrap=True, oob_score=True) 
bagging_clf.fit(X, y)
'''
CPU times: user 1.81 s, sys: 27.2 ms, total: 1.84 s
Wall time: 2.95 s
'''

---

%%time
bagging_clf = BaggingClassifier(DecisionTreeClassifier(),
                               n_estimators=500, max_samples=100,
                               bootstrap=True, oob_score=True,
                               n_jobs=-1)  # -1 所有的核
bagging_clf.fit(X, y)  
'''
CPU times: user 385 ms, sys: 56.1 ms, total: 441 ms
Wall time: 1.83 s
'''

---

随机采样 bootstrap_features

# max_features 每次取多少个特征 
random_subspaces_clf = BaggingClassifier(DecisionTreeClassifier(),
                               n_estimators=500, max_samples=500,  # 只有500个样本，取500个；相当于没有对样本数据随机
                               bootstrap=True, oob_score=True,
                               max_features=1, bootstrap_features=True)
random_subspaces_clf.fit(X, y)
random_subspaces_clf.oob_score_  # 0.83399999999999996

random_patches_clf = BaggingClassifier(DecisionTreeClassifier(),
                               n_estimators=500, max_samples=100, # 既 对样本数进行有放回随机采样，又对 特征进行...
                               bootstrap=True, oob_score=True,
                               max_features=1, bootstrap_features=True)
random_patches_clf.fit(X, y)
random_patches_clf.oob_score_  # 0.85799999999999998

---

随机森林

from sklearn.ensemble import RandomForestClassifier

rf_clf = RandomForestClassifier(n_estimators=500, oob_score=True, random_state=666, n_jobs=-1)
rf_clf.fit(X, y)

RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',
            max_depth=None, max_features='auto', max_leaf_nodes=None,
            min_impurity_decrease=0.0, min_impurity_split=None,
            min_samples_leaf=1, min_samples_split=2,
            min_weight_fraction_leaf=0.0, n_estimators=500, n_jobs=-1,
            oob_score=True, random_state=666, verbose=0, warm_start=False)

rf_clf.oob_score_ # 0.89200000000000002

---

rf_clf2 = RandomForestClassifier(n_estimators=500, max_leaf_nodes=16, oob_score=True, random_state=666, n_jobs=-1)
rf_clf2.fit(X, y)
rf_clf2.oob_score_  # 0.90600000000000003

随机森林拥有决策树和BaggingClassifier的所有参数

---

Extra-Trees

from sklearn.ensemble import ExtraTreesClassifier

et_clf = ExtraTreesClassifier(n_estimators=500, bootstrap=True, oob_score=True, random_state=666, n_jobs=-1)
et_clf.fit(X, y)

ExtraTreesClassifier(bootstrap=True, class_weight=None, criterion='gini',
           max_depth=None, max_features='auto', max_leaf_nodes=None,
           min_impurity_decrease=0.0, min_impurity_split=None,
           min_samples_leaf=1, min_samples_split=2,
           min_weight_fraction_leaf=0.0, n_estimators=500, n_jobs=-1,
           oob_score=True, random_state=666, verbose=0, warm_start=False)

et_clf.oob_score_  # 0.89200000000000002

---

集成学习解决回归问题

from sklearn.ensemble import BaggingRegressor
from sklearn.ensemble import RandomForestRegressor
from sklearn.ensemble import ExtraTreesRegressor

Boosting

AdaBoosting

不再有 oob，还是使用 train_test_split。

from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import AdaBoostClassifier

ada_clf = AdaBoostClassifier(DecisionTreeClassifier(max_depth=2), n_estimators=500)
ada_clf.fit(X_train, y_train)

AdaBoostClassifier(algorithm='SAMME.R',
          base_estimator=DecisionTreeClassifier(class_weight=None, criterion='gini', max_depth=2,
            max_features=None, max_leaf_nodes=None,
            min_impurity_decrease=0.0, min_impurity_split=None,
            min_samples_leaf=1, min_samples_split=2,
            min_weight_fraction_leaf=0.0, presort=False, random_state=None,
            splitter='best'),
          learning_rate=1.0, n_estimators=500, random_state=None)

ada_clf.score(X_test, y_test)  # 0.85599999999999998

---

Gradient Boosting

from sklearn.ensemble import GradientBoostingClassifier

gb_clf = GradientBoostingClassifier(max_depth=2, n_estimators=30)
gb_clf.fit(X_train, y_train)

GradientBoostingClassifier(criterion='friedman_mse', init=None,
              learning_rate=0.1, loss='deviance', max_depth=2,
              max_features=None, max_leaf_nodes=None,
              min_impurity_decrease=0.0, min_impurity_split=None,
              min_samples_leaf=1, min_samples_split=2,
              min_weight_fraction_leaf=0.0, n_estimators=30,
              presort='auto', random_state=None, subsample=1.0, verbose=0,
              warm_start=False)

gb_clf.score(X_test, y_test)  # 0.90400000000000003

---

Boosting 解决回归问题

from sklearn.ensemble import AdaBoostRegressor
from sklearn.ensemble import GradientBoostingRegressor