CODEWITHSUNDEEP
pythonmachine-learningscikit-learndata-scienceensemble-learning
Querenker
Querenker

Reputation: 2365

Sklearn ensemble model with result data instead of classifiers

For example the VotingClassifier expects a list of estimators, but in my case the different estimators already produced results (in the form of probabilities for each possible label e.g. [0.8, 0.2, 0.0, 0.0]) for the traing dataset as well as the result dataset. Is there a way to use this instead of the actual classifiers?

Upvotes: -1

Views: 97

Answers (1)

Roelant
Roelant

Reputation: 5119

If you already have the probabilities calculated, than you can use a simple equivalent numpy code. Note that you would need to generalise the example to the case with many predictions :)

import numpy as np

class_1 = [0.5, 0.4, 0.1, 0.0]
class_2 = [0.0, 0.4, 0.6, 0.0] 
class_3 = [0.5, 0.4, 0.05, 0.05] 
class_combined = np.array([class_1, class_2, class_3])
class_combined

# VotingClassifier(voting='hard')
hard_voting = np_matrix.argmax(axis=1)
hard = np.bincount(voting).argmax()
0

# VotingClassifier(voting='soft')
soft_sum = class_combined.sum(axis=0)
soft = soft_sum.argmax()
1

Upvotes: 2

Related Questions