Solution contains target Y labels.

[sheng-di]

Example code:

from pymrmre import mrmr
import numpy as np
import pandas as pd

data_X = np.random.random((10, 7))
data_Y = np.random.random_integers(0, 1, (10, 1))
print(data_Y)
X = pd.DataFrame(data=data_X, columns=[f'f{i}' for i in range(7)])
Y = pd.DataFrame(data=data_Y, columns=['label'])
fixed_features = ['f6']
category_features = [f'f{i}' for i in range(6)]

solutions = mrmr.mrmr_ensemble(features=X, targets=Y, fixed_features=fixed_features,
                               category_features=category_features,
                               solution_length=4, solution_count=3)

However, there are Y labels in solutions, which is incorrect:

('label', [['f6', 'label', 'label', 'label'], ['f6', 'label', 'label', 'label'], ['f6', 'label', 'label', 'label']])

[ChristopherEeles]

Hi @378978764,

I am going to look into both your issues, but the C++ developer who wrote the code is no longer with our organization and I am not a C++ expert. So it may take some time.

I will keep you updated.

Best,
Christopher Eeles

[ChristopherEeles]

Can you confirm the version of pymrmre you are using with pip freeze please?

[ChristopherEeles]

Hi @378978764,

I have determined that the target column name only shows up in the features when len(fixed_features) + len(categorical_features) == X.shape[1].

As I suspected, the issue is occurring in the package C++ internals. I will need to ask my colleagues for help debugging that code.

For now, I am going to throw an error in the Python code when that case occurs. Look for version 1.0.6 for this fix.

Best,
Chris