How to save and reused in a file with joblib a model developed with scikit learn in python ?

Example of how to save in a file with joblib a model developed with scikit learn in python:

Generate random numbers

from sklearn import linear_model
from pylab import figure

import matplotlib.pyplot as plt
import numpy as np

x = np.random.uniform(0,8,100)

sigma = np.random.randn(100) * 4.1

y = 4.0 * x + 2.0 + sigma

fig = figure(num=None, figsize=(12, 10), dpi=80, facecolor='w', edgecolor='k')

plt.scatter(x,y)

plt.title(r'Linear regression with scikit learn in python')
plt.xlabel('x')
plt.ylabel('y')
plt.xlim(0,8)

plt.savefig("linear_regression_01.png", bbox_inches='tight')

Train a model

reg = linear_model.LinearRegression()

x = x[:, np.newaxis]
y = y[:, np.newaxis]

reg.fit(x,y)

xp = np.arange(0,8,0.2)
xp = xp[:, np.newaxis]
yp = reg.predict(xp)

plt.plot(xp,yp, color='coral')

plt.title(r'Linear regression with scikit learn in python')
plt.xlabel('x')
plt.ylabel('y')
plt.xlim(0,8)

plt.savefig("linear_regression_02.png", bbox_inches='tight')

plt.show()

Save the model in a file

from joblib import dump, load

dump(reg, 'regression_model_saved.joblib')

Load the model from the file

reg_loaded = load('regression_model_saved.joblib')

xp = np.arange(0,8,0.2)
xp = xp[:, np.newaxis]

yp = reg_loaded.predict(xp)

plt.scatter(x,y)
plt.plot(xp,yp, color='coral')

plt.title(r'Linear regression with scikit learn in python')
plt.xlabel('x')
plt.ylabel('y')
plt.xlim(0,8)

plt.show()

Save multiple models in a same file

Create two models

fig = figure(num=None, figsize=(12, 10), dpi=80, facecolor='w', edgecolor='k')

x = np.random.uniform(0,8,100)

sigma = np.random.randn(100) * 4.1

y = 4.0 * x + 2.0 + sigma

plt.scatter(x,y, color='coral')

reg = linear_model.LinearRegression()

x = x[:, np.newaxis]
y = y[:, np.newaxis]

reg1 = reg.fit(x,y)

x = np.random.uniform(0,8,100)

sigma = np.random.randn(100) * 4.1


plt.plot(xp,reg1.predict(xp), color='coral')

y = 6.0 * x - 2.0 + sigma

x = x[:, np.newaxis]
y = y[:, np.newaxis]

reg = linear_model.LinearRegression()

reg2 = reg.fit(x,y)

xp = np.arange(0,8,0.2)
xp = xp[:, np.newaxis]

plt.scatter(x,y, color='lightblue')
plt.plot(xp,reg2.predict(xp), color='lightblue')

plt.title(r'Linear regression with scikit learn in python')
plt.xlabel('x')
plt.ylabel('y')
plt.xlim(0,8)

plt.savefig("linear_regression_03.png", bbox_inches='tight')

plt.show()

Save the two models in a file:

dump([reg1, reg2], 'regression_model_saved.joblib', compress=1)

Use loaded models

reg1_loaded, reg2_loaded = load('regression_model_saved.joblib')

xp = np.arange(0,8,0.2)
xp = xp[:, np.newaxis]

plt.plot(xp,reg1_loaded.predict(xp), color='coral')
plt.plot(xp,reg2_loaded.predict(xp), color='lightblue')

plt.title(r'Linear regression with scikit learn in python')
plt.xlabel('x')
plt.ylabel('y')
plt.xlim(0,8)

plt.show()

References

• Persistence example
• Can use joblib dump many object?
• sklearn dumping model using joblib, dumps multiple files. Which one is the correct model?
• Linear Models