保存机器学习模型到本地——pickle和joblib

版权声明：本文为博主原创文章，未经博主允许不得转载。 https://blog.csdn.net/Maverick_7/article/details/82355245

from sklearn.linear_model import LogisticRegression  
from sklearn.datasets import load_iris  
from sklearn.model_selection import train_test_split

# Load and split data
data = load_iris()  
Xtrain, Xtest, Ytrain, Ytest = train_test_split(data.data, data.target, test_size=0.3, random_state=4)

# Create a model
model = LogisticRegression(C=0.1,  
                           max_iter=20, 
                           fit_intercept=True, 
                           n_jobs=3, 
                           solver='liblinear')
model.fit(Xtrain, Ytrain) 

"""
our resulting model:
LogisticRegression(C=0.1, class_weight=None, dual=False, fit_intercept=True,  
    intercept_scaling=1, max_iter=20, multi_class='ovr', n_jobs=3,
    penalty='l2', random_state=None, solver='liblinear', tol=0.0001,
    verbose=0, warm_start=False)
"""

Pickle Module (also: cPickle)

导入pickle模块保存模型到本地：

import pickle

#
# Create your model here (same as above)
#

# Save to file in the current working directory
pkl_filename = "pickle_model.pkl"  
with open(pkl_filename, 'wb') as file:  
    pickle.dump(model, file)

# Load from file
with open(pkl_filename, 'rb') as file:  
    pickle_model = pickle.load(file)

# Calculate the accuracy score and predict target values
score = pickle_model.score(Xtest, Ytest)  
print("Test score: {0:.2f} %".format(100 * score))  
Ypredict = pickle_model.predict(Xtest)  

也可以将一些过程中的参数通过tuple的形式保存下来：

tuple_objects = (model, Xtrain, Ytrain, score)

# Save tuple
pickle.dump(tuple_objects, open("tuple_model.pkl", 'wb'))

# Restore tuple
pickled_model, pickled_Xtrain, pickled_Ytrain, pickled_score = pickle.load(open("tuple_model.pkl", 'rb'))

Joblib Module

from sklearn.externals import joblib

# Save to file in the current working directory
joblib_file = "joblib_model.pkl"  
joblib.dump(model, joblib_file)

# Load from file
joblib_model = joblib.load(joblib_file)

# Calculate the accuracy and predictions
score = joblib_model.score(Xtest, Ytest)  
print("Test score: {0:.2f} %".format(100 * score))  
Ypredict = pickle_model.predict(Xtest)  

也可以自己DIY存储模块，Manual Save and Restore to JSON

import json  
import numpy as np

class MyLogReg(LogisticRegression):

    # Override the class constructor
    def __init__(self, C=1.0, solver='liblinear', max_iter=100, X_train=None, Y_train=None):
        LogisticRegression.__init__(self, C=C, solver=solver, max_iter=max_iter)
        self.X_train = X_train
        self.Y_train = Y_train

    # A method for saving object data to JSON file
    def save_json(self, filepath):
        dict_ = {}
        dict_['C'] = self.C
        dict_['max_iter'] = self.max_iter
        dict_['solver'] = self.solver
        dict_['X_train'] = self.X_train.tolist() if self.X_train is not None else 'None'
        dict_['Y_train'] = self.Y_train.tolist() if self.Y_train is not None else 'None'

        # Creat json and save to file
        json_txt = json.dumps(dict_, indent=4)
        with open(filepath, 'w') as file:
            file.write(json_txt)

    # A method for loading data from JSON file
    def load_json(self, filepath):
        with open(filepath, 'r') as file:
            dict_ = json.load(file)

        self.C = dict_['C']
        self.max_iter = dict_['max_iter']
        self.solver = dict_['solver']
        self.X_train = np.asarray(dict_['X_train']) if dict_['X_train'] != 'None' else None
        self.Y_train = np.asarray(dict_['Y_train']) if dict_['Y_train'] != 'None' else None

应用模块：

filepath = "mylogreg.json"

# Create a model and train it
mylogreg = MyLogReg(X_train=Xtrain, Y_train=Ytrain)  
mylogreg.save_json(filepath)

# Create a new object and load its data from JSON file
json_mylogreg = MyLogReg()  
json_mylogreg.load_json(filepath)  
json_mylogreg  


"""
MyLogReg(C=1.0,  
     X_train=array([[ 4.3,  3. ,  1.1,  0.1],
       [ 5.7,  4.4,  1.5,  0.4],
       ...,
       [ 7.2,  3. ,  5.8,  1.6],
       [ 7.7,  2.8,  6.7,  2. ]]),
     Y_train=array([0, 0, ..., 2, 2]), class_weight=None, dual=False,
     fit_intercept=True, intercept_scaling=1, max_iter=100,
     multi_class='ovr', n_jobs=1, penalty='l2', random_state=None,
     solver='liblinear', tol=0.0001, verbose=0, warm_start=False)
"""

Save and Restore Models

Saving machine learning models

Model persistence