Некоторые учебные записи об обучении в ансамбле

машинное обучение

Прежде всего, я провел три или четыре дня на прошлой неделе, чтобы посмотреть некоторые знания об обучении ансамблю, Принцип не был изучен глубоко, я понял идею интегрированного обучения и некоторые принципы реализации на макроуровне. .

  • В основном изучил два вида параллельной упаковки моделей и последовательного бустинга моделей, а также укладки в стопку.
  • Существует три основных стратегии интеграции: среднее голосование и обучающий объединитель.

1 Интегрированные рукописные заметки для обучения

Конкретные примечания прилагаются ниже:在这里插入图片描述 在这里插入图片描述 在这里插入图片描述 在这里插入图片描述

2 Ensemble Learning Demo

2.1 Voting

from sklearn import datasets, model_selection
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression
from sklearn.neighbors import KNeighborsClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import VotingClassifier
from sklearn.svm import SVC
from sklearn.metrics import accuracy_score

digits = datasets.load_digits()
x_train, x_test, y_train, y_test = train_test_split(digits.data, digits.target, train_size=0.7, random_state=1)

print(">>>VotingClassifierTest<<<")
knn = KNeighborsClassifier(n_neighbors=1)
dtree = DecisionTreeClassifier()
lr = LogisticRegression(max_iter=10000)
svm = SVC()
voting_clf = VotingClassifier([('knn', knn), ('dt', dtree), ('lr', lr), ('svm', svm)])

for clf in (knn, dtree, lr, svm, voting_clf):
    clf.fit(x_train, y_train)
    print(clf.__class__.__name__, "accuracy: %0.5f%%" % (clf.score(x_test, y_test) * 100))
    #scores = model_selection.cross_val_score(clf, x_data, y_data, cv=5, scoring='accuracy')
    #print(clf.__class__.__name__, "acc = %0.5f" % (scores.mean()))

# for clf in (knn, dt, lr, svm, voting_clf):
#     clf.fit(x_data, y_data)
#     y_predict = clf.predict(x_data)
#     print(clf.__class__.__name__, "acc = %0.5f" % accuracy_score(y_data, y_predict))

在这里插入图片描述

2.2 Bagging

from sklearn import datasets
from sklearn.ensemble import BaggingClassifier
from sklearn import tree
from sklearn.model_selection import train_test_split

digits = datasets.load_digits()
x_train, x_test, y_train, y_test = train_test_split(digits.data, digits.target, train_size=0.75, random_state=55)

print(">>>BaggingClassifierTest<<<")
dtree = tree.DecisionTreeClassifier()
bagging_tree = BaggingClassifier(base_estimator=dtree, n_estimators=800) # 使用bagging算法训练多组决策树分类器

dtree.fit(x_train, y_train)
print("DecisionTree accuracy: %0.5f%%" % (dtree.score(x_test, y_test) * 100))

bagging_tree.fit(x_train, y_train)
print("BaggingTree accuracy: %0.5f%%" % (bagging_tree.score(x_test, y_test) * 100))

在这里插入图片描述

2.3 AdaBoost

from sklearn import datasets
from sklearn.ensemble import AdaBoostClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.model_selection import train_test_split
# 这里由于不在采用随机抽样，因此需要分割数据集
digits = datasets.load_digits()
x_train, x_test, y_train, y_test = train_test_split(digits.data, digits.target, train_size=0.7, random_state=88)

print(">>>AdaBoostClassifierTest<<<")

# 这里根据决策树的调参方式进行调参
dtree = DecisionTreeClassifier()
adaboost_clf = AdaBoostClassifier(base_estimator=DecisionTreeClassifier(max_depth=2),
                                  n_estimators=600, learning_rate=0.5, algorithm="SAMME")

dtree.fit(x_train, y_train)
print("DecisionTree accuracy: %0.5f%%" % (dtree.score(x_test, y_test) * 100))

adaboost_clf.fit(x_train, y_train)
print("AdaBoost accuracy: %0.5f%%" % (adaboost_clf.score(x_test, y_test) * 100))

在这里插入图片描述

2.4 XGBoost

from sklearn import datasets
from sklearn.model_selection import train_test_split
from xgboost import XGBClassifier
from sklearn.metrics import accuracy_score
from sklearn.tree import DecisionTreeClassifier

digits = datasets.load_digits()
x_train, x_test, y_train, y_test = train_test_split(digits.data, digits.target, test_size=0.3, random_state=33)

print(">>>XGBoostClassifierTest<<<")

xgb = XGBClassifier()
dt = DecisionTreeClassifier()

dt.fit(x_train, y_train)
y_predict = dt.predict(x_test)
accuracy = accuracy_score(y_test, y_predict)
print("DecisionTree accuarcy: %.5f%%" % (accuracy * 100.0))

xgb.fit(x_train, y_train)
y_predict = xgb.predict(x_test)
accuracy = accuracy_score(y_test, y_predict)
print("XGBoost accuarcy: %.5f%%" % (accuracy * 100.0))

# 特征重要性
# import matplotlib.pyplot as plt
# from xgboost import plot_importance
# fig,ax = plt.subplots(figsize=(10,15))
# plot_importance(xgb, height=0.5, max_num_features=64, ax=ax)
# plt.show()

在这里插入图片描述

2.5 CatBoost

rom catboost import CatBoostClassifier, Pool
from sklearn import datasets
from sklearn.model_selection import train_test_split
import numpy as np

train_data = Pool(data=[[1, 4, 5, 6],
                        [4, 5, 6, 7],
                        [30, 40, 50, 60]],
                  label=[1, 1, -1],
                  weight=[0.1, 0.2, 0.3])

model = CatBoostClassifier(iterations=10)

model.fit(train_data, plot=True)
preds_class = model.predict(train_data)
print("%0.5f%%" % (model.score(train_data) * 100))

# import matplotlib.pyplot as plt
# fea_ = model.feature_importances_
# fea_name = model.feature_names_
# plt.figure(figsize=(10, 10))
# plt.barh(fea_name,fea_,height =0.5)

# plt.show()

在这里插入图片描述Примечание. CatBoost поддерживает рисование обучающих графиков и должен запускаться в Jupyter Notebook. Ниже приведен код, который небрежно запускал набор данных Titanic, потому что CatBoost больше подходит для такого рода многофункционального обучения и высококачественного прогнозирования без настройки параметров. Но мой код очень грубый и точность очень низкая... У меня есть время его оптимизировать.

from catboost.datasets import titanic
from catboost import CatBoostClassifier, Pool
from sklearn.model_selection import train_test_split
import numpy as np

data_train, data_test = titanic()
print(data_train.head(10))

null_value_stats = data_train.isnull().sum(axis=0)
print(null_value_stats[null_value_stats != 0])
data_train.fillna(-999, inplace=True)
data_test.fillna(-999, inplace=True)
X = data_train.drop('Survived', axis=1)
y = data_train.Survived
print("data:", X)
print("label:", y)
print(X.dtypes)

# remove = ['PassengerId', 'Cabin', 'Name', 'Ticket']
# X = data_train.drop(remove, axis=1)
# X = data_train.dropna(how='any', axis='rows')
# y = X.pop('Survived')
# print(X.head(10))
#
x_train, x_test, y_train, y_test = train_test_split(X, y, train_size=0.7, random_state=66)
cat_features = np.where(X.dtypes != float)[0]
# cat_features = [0, 1, 6]
train_pool = Pool(x_train, y_train, cat_features=cat_features)
test_pool = Pool(x_test, y_test, cat_features=cat_features)

cat_clf = CatBoostClassifier()
cat_clf.fit(train_pool, eval_set=test_pool, plot=True, silent=False, use_best_model=True)

2.6 Stacking

from sklearn import datasets, model_selection
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression
from sklearn.neighbors import KNeighborsClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import StackingClassifier
from sklearn.svm import SVC
from sklearn.metrics import accuracy_score

digits = datasets.load_digits()
print(digits)
x_train, x_test, y_train, y_test = train_test_split(digits.data, digits.target, train_size=0.7, random_state=66)

print(">>>StackingClassifierTest<<<")

knn = KNeighborsClassifier(n_neighbors=1)
dtree = DecisionTreeClassifier()
lr = LogisticRegression(max_iter=10000)
svm = SVC()
stacking_clf = StackingClassifier(estimators=[('knn', knn), ('dt', dtree), ('lr', lr)], final_estimator=svm)

for clf in (knn, dtree, lr, svm, stacking_clf):
    clf.fit(x_train, y_train)
    print(clf.__class__.__name__, "accuracy: %0.5f%%" % (clf.score(x_test, y_test) * 100))
    # scores = model_selection.cross_val_score(clf, x_data, y_data, cv=5, scoring='accuracy')
    # print(clf.__class__.__name__, "acc = %0.5f" % (scores.mean()))

在这里插入图片描述

3 ссылки

Категории