长话短说stacking集成学习算法，保证你能看得懂（3)

这是在乳腺癌数据集的的stacking集成学习案例

1. 读入样本数据

import warnings

warnings.filterwarnings(“ignore”)
from sklearn import datasets
file=datasets.load_breast_cancer()
X=file[“data”]
y=file[“target”]
#拿出20%，用作测试数据

2. 分拆为训练集和测试集

from sklearn.model_selection import train_test_split
X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.2)

3. 初级学习——通过交叉验证产生次级学习数据集

import numpy as np
from sklearn.model_selection import KFold
def stacking(model, x_train, y_train, x_test, k=5):

    #这个函数是stacking的核心，使用交叉验证的方法得到第二阶段训练集和测试集

    #取训练集和测试集的记录数
    train_num, test_num = x_train.shape[0], x_test.shape[0]
    
    #初始化训练集和测试集
    second_level_train_set = np.zeros(train_num)
    second_level_test_set = np.zeros(test_num)
    #初始化每轮在测试集上预测的结果
    test_nfolds_sets = np.zeros((test_num, k))
    
    kf = KFold(n_splits=k)

    #在训练集上进行k折交叉验证
    for i,(train_index, test_index) in enumerate(kf.split(x_train)):
        x_tr, y_tr = x_train[train_index], y_train[train_index]
        x_ts=  x_train[test_index]

        #第i轮训练
        model.fit(x_tr, y_tr)

        #第i轮在训练集上产生的测试结果
        second_level_train_set[test_index] = clf.predict(x_ts)
       
        #第i轮模型在测试集上产生的测试结果
        test_nfolds_sets[:,i] = model.predict(x_test)
    
    #对k轮产生的测试集上的预测结果进行算术平均，拟合成模型clf在测试集的预测结果y_test_i
    second_level_test_set[:] = np.round(test_nfolds_sets.mean(axis=1))
    #输出clf模型在训练集和测试集上的预测结果
    return second_level_train_set, second_level_test_set

#初始化融合后的训练集和测试集
train_sets = []
test_sets = []

from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier,GradientBoostingClassifier
from sklearn.svm import SVC
from xgboost import XGBClassifier
from lightgbm import LGBMClassifier

rf= RandomForestClassifier()
ada = AdaBoostClassifier()
gbdt = GradientBoostingClassifier()
svc= SVC()
lgbm=LGBMClassifier()
xgb=XGBClassifier()

for clf in [rf,ada,gbdt,svc,lgbm,xgb]:
    train_set, test_set = stacking(clf, X_train,y_train,X_test)
    train_sets.append(train_set)
    test_sets.append(test_set)

#将每个学习器的预测结果列化，并联，形成n维向量数据集，构造新的训练集和测试集
X_train_new = np.concatenate([result_set.reshape(-1,1) for result_set in train_sets], axis=1)
X_test_new = np.concatenate([y_test_set.reshape(-1,1) for y_test_set in test_sets], axis=1)

4. 次级学习及预测

#我们这里使用6个分类算法。正常来说，这六个分类器是通过参数调优的，在此略过

#使用lightGBM作为最终的学习器
lgbm.fit(X_train_new, y_train)
y_pred= lgbm.predict(X_test_new)
from sklearn.metrics import accuracy_score
from sklearn.metrics import classification_report
from sklearn.metrics import mean_squared_error,confusion_matrix
accuracy_score(y_test,y_pred)

推荐自编课程《零基础学python数据分析》
推荐自编简明预测分析教程(Python版)