XGBoost can be used to create some of the most performant models for tabular data using the gradient boosting algorithm.
Once trained, it is often a good practice to save your model to file for later use in making predictions new test and validation datasets and entirely new data.
In this post you will discover how to save your XGBoost models to file using the standard Python pickle API.
After completing this tutorial, you will know:
- How to save and later load your trained XGBoost model using pickle.
- How to save and later load your trained XGBoost model using joblib.
Kick-start your project with my new book XGBoost With Python, including step-by-step tutorials and the Python source code files for all examples.
Let’s get started.
- Update Jan/2017: Updated to reflect changes in scikit-learn API version 0.18.1.
- Update Mar/2018: Added alternate link to download the dataset as the original appears to have been taken down.
- Update Oct/2019: Updated to use Joblib API directly.

How to Save Gradient Boosting Models with XGBoost in Python
Photo by Keoni Cabral, some rights reserved.
Need help with XGBoost in Python?
Take my free 7-day email course and discover xgboost (with sample code).
Click to sign-up now and also get a free PDF Ebook version of the course.
Start Your FREE Mini-Course Now!
Serialize Your XGBoost Model with Pickle
Pickle is the standard way of serializing objects in Python.
You can use the Python pickle API to serialize your machine learning algorithms and save the serialized format to a file, for example:
# save model to file pickle.dump(model, open("pima.pickle.dat", "wb")) |
Later you can load this file to deserialize your model and use it to make new predictions, for example:
# load model from file loaded_model = pickle.load(open("pima.pickle.dat", "rb")) |
The example below demonstrates how you can train a XGBoost model on the Pima Indians onset of diabetes dataset, save the model to file and later load it to make predictions.
Download the dataset and save it to your current working directory.
The full code listing is provided below for completeness.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 |
# Train XGBoost model, save to file using pickle, load and make predictions from numpy import loadtxt import xgboost import pickle from sklearn import model_selection from sklearn.metrics import accuracy_score # load data dataset = loadtxt('pima-indians-diabetes.csv', delimiter=",") # split data into X and y X = dataset[:,0:8] Y = dataset[:,8] # split data into train and test sets seed = 7 test_size = 0.33 X_train, X_test, y_train, y_test = cross_validation.train_test_split(X, Y, test_size=test_size, random_state=seed) # fit model no training data model = xgboost.XGBClassifier() model.fit(X_train, y_train) # save model to file pickle.dump(model, open("pima.pickle.dat", "wb"))
# some time later...
# load model from file loaded_model = pickle.load(open("pima.pickle.dat", "rb")) # make predictions for test data y_pred = loaded_model.predict(X_test) predictions = [round(value) for value in y_pred] # evaluate predictions accuracy = accuracy_score(y_test, predictions) print("Accuracy: %.2f%%" % (accuracy * 100.0)) |
Running this example saves your trained XGBoost model to the pima.pickle.dat pickle file in the current working directory.
Note: Your results may vary given the stochastic nature of the algorithm or evaluation procedure, or differences in numerical precision. Consider running the example a few times and compare the average outcome.
After loading the model and making predictions on the training dataset, the accuracy of the model is printed.
Serialize XGBoost Model with joblib
Joblib is part of the SciPy ecosystem and provides utilities for pipelining Python jobs.
The Joblib API provides utilities for saving and loading Python objects that make use of NumPy data structures, efficiently. It may be a faster approach for you to use with very large models.
The API looks a lot like the pickle API, for example, you may save your trained model as follows:
# save model to file joblib.dump(model, "pima.joblib.dat") |
You can later load the model from file and use it to make predictions as follows:
# load model from file loaded_model = joblib.load("pima.joblib.dat") |
The example below demonstrates how you can train an XGBoost model for classification on the Pima Indians onset of diabetes dataset, save the model to file using Joblib and load it at a later time in order to make predictions.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 |
# Train XGBoost model, save to file using joblib, load and make predictions from numpy import loadtxt from xgboost import XGBClassifier from joblib import dump from joblib import load from sklearn.model_selection import train_test_split from sklearn.metrics import accuracy_score # load data dataset = loadtxt('pima-indians-diabetes.csv', delimiter=",") # split data into X and y X = dataset[:,0:8] Y = dataset[:,8] # split data into train and test sets seed = 7 test_size = 0.33 X_train, X_test, y_train, y_test = train_test_split(X, Y, test_size=test_size, random_state=seed) # fit model on training data model = XGBClassifier() model.fit(X_train, y_train) # save model to file dump(model, "pima.joblib.dat") print("Saved model to: pima.joblib.dat")
# some time later...
# load model from file loaded_model = load("pima.joblib.dat") print("Loaded model from: pima.joblib.dat") # make predictions for test data predictions = loaded_model.predict(X_test) # evaluate predictions accuracy = accuracy_score(y_test, predictions) print("Accuracy: %.2f%%" % (accuracy * 100.0)) |
Running the example saves the model to file as pima.joblib.dat in the current working directory and also creates one file for each NumPy array within the model (in this case two additional files).
pima.joblib.dat pima.joblib.dat_01.npy pima.joblib.dat_02.npy |
Note: Your results may vary given the stochastic nature of the algorithm or evaluation procedure, or differences in numerical precision. Consider running the example a few times and compare the average outcome.
After the model is loaded, it is evaluated on the training dataset and the accuracy of the predictions is printed.
Summary
In this post, you discovered how to serialize your trained XGBoost models and later load them in order to make predictions.
Specifically, you learned:
- How to serialize and later load your trained XGBoost model using the pickle API.
- How to serialize and later load your trained XGBoost model using the joblib API.
Do you have any questions about serializing your XGBoost models or about this post? Ask your questions in the comments and I will do my best to answer.