scikit learn 模块 调参 pipeline+girdsearch 数据举例来说:文档分类 (python代码)

scikit learn 模块 调参 pipeline+girdsearch 数据举例:文档分类 (python代码)
scikit learn 模块 调参 pipeline+girdsearch 数据举例:文档分类数据集 fetch_20newsgroups

 

#-*- coding: UTF-8 -*-

import numpy as np
from sklearn.pipeline import Pipeline
from sklearn.linear_model import SGDClassifier
from sklearn.grid_search import GridSearchCV
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.feature_extraction.text import TfidfTransformer
from sklearn.datasets import fetch_20newsgroups
from sklearn import metrics

获取待分类的文本数据源
categories = ['comp.graphics', 'comp.os.ms-windows.misc','comp.sys.ibm.pc.hardware','comp.sys.mac.hardware','comp.windows.x'];
newsgroup_data = fetch_20newsgroups(subset = 'train',categories = categories)
X,Y=np.array(newsgroup_data.data),np.array(newsgroup_data.target)
Xtrain,Ytrain,Xtest,Ytest =X[0:2400],Y[0:2400],X[2400:],Y[2400:]


#Pipeline主要用于将三个需要串行的模块串在一起,后一个模型处理前一个的结果'''
#vect主要用于去音调、转小写、去停顿词->tdidf主要用于计词频->clf分类模型'''
pipeline_obj = Pipeline([('vect',CountVectorizer()),('tfidf',TfidfTransformer()),('clf',SGDClassifier()),])
print "pipeline:",'n', [name for name, _ in pipeline_obj.steps],'n'


#定义需要遍历的所有候选参数的字典,key_name需要用__分隔模型名和模型内部的参数名'''
parameters = {
    'vect__max_df': (0.5, 0.75),'vect__max_features': (None, 5000, 10000),
    'tfidf__use_idf': (True, False),'tfidf__norm': ('l1', 'l2'),
    'clf__alpha': (0.00001, 0.000001), 'clf__n_iter': (10, 50) }
print "parameters:",'n',parameters,'n'


#GridSearchCV用于寻找vectorizer词频统计, tfidftransformer特征变换和SGD classifier分类模型的最优参数
grid_search = GridSearchCV( pipeline_obj, parameters, n_jobs = 1,verbose=1 )
print 'grid_search','n',grid_search,'n' #输出所有参数名及参数候选值
grid_search.fit(Xtrain,Ytrain),'n'#遍历执行候选参数,寻找最优参数

best_parameters = dict(grid_search.best_estimator_.get_params())#get实例中的最优参数
for param_name in sorted(parameters.keys()):
    print("t%s: %r" % (param_name, best_parameters[param_name])),'n'#输出最有参数结果
pipeline_obj.set_params(clf__alpha = 1e-05,clf__n_iter = 50,tfidf__use_idf = True,vect__max_df = 0.5,vect__max_features = None)
#将pipeline_obj实例中的参数重写为最优结果'''
print pipeline_obj.named_steps


#用最优参数训练模型'''
pipeline_obj.fit(Xtrain,Ytrain)
pred = pipeline_obj.predict(Xtrain)
print 'n',metrics.classification_report(Ytrain,pred)
pred = pipeline_obj.predict(Xtest)
print 'n',metrics.classification_report(Ytest,pred)

执行结果:总共有96个参数排列组合候选组,每组跑3次模型进行交叉验证,共计跑模型96*3=288次。

 

调参前VS调参后:

#参考

#http://blog.csdn.net/mmc2015/article/details/46991465
# http://blog.csdn.net/abcjennifer/article/details/23884761
# http://scikit-learn.org/stable/modules/pipeline.html
# http://blog.csdn.net/yuanyu5237/article/details/44278759

相关内容推荐