Using smote_variants in Julia

Similarly to R using reticulate, Python packages can be called from Julia using the package PyCall given that some python installation with smote_variants is available.

Suppose, there is an Anaconda3 install available at ‘/home/<user>/anaconda3’ and smote_variants and imbalanced_databases are installed on the base conda environment.

The following steps are needed to run oversampler codes from the smote_variants package:

  • Start the Julia interpreter:

julia

  • In the Julia prompt, set the Python path to that of the Anaconda install:

ENV["PYTHON"]= "/home/<user>/anaconda3/bin/python3"

  • Add the PyCall package:

import Pkg

Pkg.add("PyCall")

Pkg.build("PyCall")

  • Restart the Julia interpreter.

  • The following code should work:

using PyCall

@pyimport imbalanced_databases as imbd
@pyimport smote_variants as sv

dataset= imbd.load_iris0()
oversampler= sv.SMOTE_ENN()

X_samp, y_samp= oversampler[:sample](dataset["data"], dataset["target"])