How to save & load xgboost model?

Saving and Loading XGBoost Models

XGBoost models are powerful and often require significant training time. Saving a trained model allows you to reuse it without retraining, saving valuable time and resources. This article guides you through the process of saving and loading XGBoost models in Python.

Methods for Saving and Loading

There are two primary methods for saving and loading XGBoost models:

1. Using the `save_model` and `load_model` functions

• This method is straightforward and recommended for most use cases.
• It directly saves and loads the model object.

Saving the model

import xgboost as xgb
# ... (Your model training code)
xgb.save_model(model, 'xgb_model.json')

Loading the model

loaded_model = xgb.Booster()
loaded_model.load_model('xgb_model.json')

2. Using Pickle

• Pickle allows you to save Python objects, including XGBoost models.
• It’s a versatile option for serialization.

Saving the model

import pickle
# ... (Your model training code)
pickle.dump(model, open('xgb_model.pkl', 'wb'))

Loading the model

loaded_model = pickle.load(open('xgb_model.pkl', 'rb'))

Choosing the Right Method

The best method depends on your specific needs:

Example

import xgboost as xgb
import pandas as pd

# Sample data
data = pd.DataFrame({
    'feature1': [1, 2, 3, 4, 5],
    'feature2': [2, 4, 6, 8, 10],
    'target': [3, 6, 9, 12, 15]
})

# Split data into train and test
X_train = data[['feature1', 'feature2']]
y_train = data['target']

# Train XGBoost model
model = xgb.XGBRegressor()
model.fit(X_train, y_train)

# Save the model using save_model
xgb.save_model(model, 'xgb_model.json')

# Load the model
loaded_model = xgb.Booster()
loaded_model.load_model('xgb_model.json')

# Predict using the loaded model
X_test = pd.DataFrame({
    'feature1': [6, 7],
    'feature2': [12, 14]
})
predictions = loaded_model.predict(xgb.DMatrix(X_test))

print(predictions)

Conclusion

Saving and loading XGBoost models is crucial for efficient model management and reuse. By using the provided methods, you can easily store your trained models and load them for future predictions or analysis.