机器学习-训练模型的保存与恢复（sklearn）

在做模型训练的时候，尤其是在训练集上做交叉验证，通常想要将模型保存下来，然后放到独立的测试集上测试，下面介绍的是Python中训练模型的保存和再使用。

 

模型保存（pickle模块和joblib模块）

使用pickle模块或者sklearn内部的joblib

一、使用pickle模块

from sklearn import svm
from sklearn import datasets
clf=svm.SVC()
iris=datasets.load_iris()
X,y=iris.data,iris.target
clf.fit(X,y)

import pickle
s=pickle.dumps(clf)
f=open('svm.txt','w')
f.write(s)
f.close()
f2=open('svm.txt','r')
s2=f2.read()
clf2=pickle.loads(s2)
clf2.score(X,y)

二、使用joblib

joblib更适合大数据量的模型，且只能往硬盘存储，不能往字符串存储

from sklearn.externals import joblib
joblib.dump(clf,'filename.pkl')
clf=joblib.load('filename.pkl')

三、具体例子如下：

scikit-learn已经有了模型持久化的操作，导入joblib即可

from sklearn.externals import joblib

模型保存

通过joblib的dump可以将模型保存到本地，clf是训练的分类器

from sklearn.linear_model import LogisticRegression
from sklearn.svm import SVC
from sklearn.externals import joblib
def test_save_model(self):
    model_save_path = "./model_save/"
    train_X = [[0, 0], [1, 1]]
    train_y = [0, 1]
    print "Start LR method ..."
    print('LR Train classifier...')
    clf = LogisticRegression()
    clf.fit(train_X, train_y)
    print "LR Model save..."
    save_path_name = model_save_path + "lr_" + "train_model.m"
    self.is_exist(model_save_path, save_path_name)
    joblib.dump(clf, save_path_name)
    clf = joblib.load(save_path_name)
    print('LR Predict...')
    pred = clf.predict_proba(train_X)
    submit_csv_name = model_save_path + "lr" + '_submission.csv'
    self.make_submission(pred[:, 0], submit_csv_name)

    print "Start SVM method ..."
    # 训练
    print('SVM Train classifier...')
    from sklearn import svm
    clf = svm.SVC()
    clf.fit(train_X, train_y)
    # 保存
    print "SVM Model save..."
    save_path_name=model_save_path+"svm_"+"train_model.m"
    self.is_exist(model_save_path,save_path_name)
    joblib.dump(clf, save_path_name)
    clf = joblib.load(save_path_name)
    # 预测
    print('SVM Predict...')
    pred=clf.predict(train_X)
    submit_csv_name = model_save_path+"svm" + '_submission.csv'
    self.make_submission(pred, submit_csv_name)

    train_X = [[0, 1], [1, 1]]
    train_y = [0, 1]
    print clf.score(train_X, train_y, sample_weight=None)

模型从本地调回

clf = joblib.load("train_model.m")

通过joblib的load方法，加载保存的模型。

然后就可以在测试集上测试了

clf.predict(test_X) #此处test_X为特征集

参考: https://blog.csdn.net/Dream_angel_Z/article/details/47175373