管道模型(Pipeline)

#导入数据集生成器
from sklearn.datasets import make_blobs
#导入数据集拆分工具
from sklearn.model_selection import train_test_split
#导入预处理工具
from sklearn.preprocessing import StandardScaler
#导入多层感知器神经网络
from sklearn.neural_network import MLPClassifier
#导入画图工具
import matplotlib.pyplot as plt

#生成样本数量200,分类为2,标准差为5的数据集
X,y = make_blobs(n_samples=200,centers=2,cluster_std=5)
#将数据集拆分为训练集和测试集
X_train,X_test,y_train,y_test = train_test_split(X,y,random_state=38)
#对数据进行预处理
scaler = StandardScaler().fit(X_train)
X_train_scaled = scaler.transform(X_train)
X_test_scaled = scaler.transform(X_test)
#将处理后的数据形态进行打印
print(‘\n\n\n‘)
print(‘代码运行结果‘)
print(‘====================================\n‘)
#将处理后的数据形态进行打印
print(‘训练数据集:{}‘.format(X_train_scaled.shape),‘ 标签形态:{}‘.format(X_test_scaled.shape))
print(‘\n====================================‘)
print(‘\n\n\n‘)

#训练原始数据集
plt.scatter(X_train[:,0],X_train[:,1])
#经过预处理的训练集
plt.scatter(X_train_scaled[:,0],X_train_scaled[:,1],marker=‘^‘,edgecolor=‘k‘)
#添加图题
plt.title(‘training set & scaled training set‘)
#显示图片
plt.show()

#导入网格搜索
from sklearn.model_selection import GridSearchCV
#设定网格搜索的模型参数字典
params = {‘hidden_layer_sizes‘:[(50,),(100,),(100,100)],‘alpha‘:[0.0001,0.01,0.1]}
#建立网格搜索模型
grid = GridSearchCV(MLPClassifier(max_iter=1600,random_state=38),param_grid=params,cv=3,iid=False)
#拟合数据
grid.fit(X_train_scaled,y_train)
#将结果进行打印
print(‘\n\n\n‘)
print(‘代码运行结果‘)
print(‘====================================\n‘)
print(‘模型最佳得分:{:.2f}‘.format(grid.best_score_),‘ 模型最佳参数:{}‘.format(grid.best_params_))
print(‘\n====================================‘)
print(‘\n\n\n‘)

#打印模型在测试集中的得分
print(‘\n\n\n‘)
print(‘代码运行结果‘)
print(‘====================================\n‘)
print(‘测试集得分:{}‘.format(grid.score(X_test_scaled,y_test)))
print(‘\n====================================‘)
print(‘\n\n\n‘)

#导入管道模型
from sklearn.pipeline import Pipeline
#建立包含预处理和神经网络的管道模型
pipeline = Pipeline([(‘scaler‘,StandardScaler()),(‘mlp‘,MLPClassifier(max_iter=1600,random_state=38))])
#用管道模型对训练集进行拟合
pipeline.fit(X_train,y_train)
#打印管道模型的分数
print(‘使用管道模型的MLP模型评分:{:.2f}‘.format(pipeline.score(X_test,y_test)))

#设置参数字典--------(mlp__是用于指定pipeline中的mlp算法)
params = {‘mlp__hidden_layer_sizes‘:[(50,),(100,),(100,100)],‘mlp__alpha‘:[0.0001,0.001,0.01,0.1]}
#建立包含预处理和神经网络的管道模型
pipeline = Pipeline([(‘scaler‘,StandardScaler()),(‘mlp‘,MLPClassifier(max_iter=1600,random_state=38))])
#将管道模型加入网格搜索
grid = GridSearchCV(pipeline,param_grid=params,cv=3,iid=False)
#对训练集进行拟合
grid.fit(X_train,y_train)
#打印模型交叉验证分数.最佳参数和测试集得分
print(‘\n\n\n‘)
print(‘代码运行结果‘)
print(‘====================================\n‘)
print(‘交叉验证最高分:{:.2f}‘.format(grid.best_score_))
print(‘模型最有参数:{}‘.format(grid.best_params_))
print(‘测试集得分:{}‘.format(grid.score(X_test,y_test)))
print(‘\n====================================‘)
print(‘\n\n\n‘)

#打印管道模型中的步骤
print(‘\n\n\n‘)
print(‘代码运行结果‘)
print(‘====================================\n‘)
print(pipeline.steps)
print(‘\n====================================‘)
print(‘\n\n\n‘)