决策树，随机森林和极限森林算法

决策树：https://www.cnblogs.com/molieren/articles/10664954.html

from sklearn import datasets
from sklearn.model_selection import train_test_split,cross_val_score
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier,ExtraTreesClassifier
import random

if __name__ == '__main__':
    datas = datasets.load_iris()
    #打散数据：x_train,80%训练的data x_test 20%测试的data ,y_train 80%训练的target,y_test 20%测试的target
    x_train,x_test,y_train,y_test = train_test_split(datas.data, datas.target,test_size=0.2, random_state=6)

    #决策树分类器
    model = DecisionTreeClassifier(criterion="entropy").fit(x_train,y_train)
    #用80%的数据训练得出模型，然后用20%去预测，验证模型是否正确
    # Y_test = model.predict(x_test)
    # print(Y_test == y_test)

    score = model.score(x_test, y_test)
    #决策树score验证匹配率
    print(score)

    #随机森林：随机森林由决策树组成，随机性体现在每棵树的数据不一样 n_estimators=10 树的棵数
    rf = RandomForestClassifier(criterion="entropy",n_estimators=128).fit(x_train, y_train)
    rfscore = rf.score(x_test, y_test)
    #随机森林score验证匹配率
    print(+rfscore)


    #极限森林是造不同的树，每棵树的根节点，叶节点的顺序是不一样的
    ex = ExtraTreesClassifier(criterion="entropy",n_estimators=128).fit(x_train, y_train)
    exs = ex.score(x_test, y_test)
    #极限森林score验证匹配率
    print(exs)

    # 验证不用score模式，用交叉验证更准确，cross_val_score  一个数据集合既可能做训练集，也可能做测试集，避免过拟合
    #cv=10 将数据分成10分，每一份做一次测试集
    model1 = DecisionTreeClassifier(criterion="entropy")
    rf1 = RandomForestClassifier(criterion="entropy", n_estimators=128)
    ex1 = ExtraTreesClassifier(criterion="entropy", n_estimators=128)
    momean1 = cross_val_score(model, x_train, y_train, cv=10).mean()
    rfmean1 = cross_val_score(rf, x_train, y_train, cv=10).mean()
    exmean1 = cross_val_score(ex, x_train, y_train, cv=10).mean()
    #决策树交叉验证
    print(momean1)
    #随机森林交叉验证
    print(rfmean1)
    #极限森林交叉验证
    print(exmean1)

    #打散数据，利用随机的下标
    model2 = DecisionTreeClassifier(criterion="entropy")
    rf2 = RandomForestClassifier(criterion="entropy", n_estimators=128)
    ex2 = ExtraTreesClassifier(criterion="entropy", n_estimators=128)
    list = [x for x in range(150)]
    random.shuffle(list)
    X=[datas.data[index] for index in list ]
    Y=[datas.target[index] for index in list]
    momean2 = cross_val_score(model, X, Y, cv=10).mean()
    rfmean2 = cross_val_score(rf, X, Y, cv=10).mean()
    exmean2 = cross_val_score(ex, X, Y, cv=10).mean()
    #打散数据后决策树交叉验证配配率
    print(momean1)
    #"打散数据后随机森林交叉验证配配率
    print(rfmean1)
    #打散数据后极限森林交叉验证配配率
    print(exmean1)