CODEWITHSUNDEEP
pythonscikit-learn
user1801745
user1801745

Reputation: 389

Scikit-learn with different outputs calculation

I used sudo apt-get install python-numpy python-scipy python-matplotlib python-sklearn for scikit-learn installation on rPi3 but an old version was applied!

This old version installed (v. 0.18) got output results complete different over same algorithms if compare with Windows or Mac installation (v. 0.20.1)

Windows/MAC results: R2: 0.992 MSE: 0.011 SEP: 0.104 RPD: 11.509

Raspberry results: R2: -19.00065 MSE: 28.60711 SEP: 5.34856 RPD: 0.22360

The same algorithm use PLSRegression.

Trying to use pip install on rPi3 bring a lot of compilation errors (no way). So, why different results? Is this because the different version?

Upvotes: 0

Views: 76

Answers (1)

Matthieu Brucher
Matthieu Brucher

Reputation: 22023

From the change log, you can see that in version 0.19, changes were made to the PLSRegression model.

This would explain the difference, as the default argument is scale=True.

Upvotes: 1

Related Questions