CODEWITHSUNDEEP
scikit-learnnaivebayes
123456
123456

Reputation: 145

Can I set a random state for the Complement Naive Bayes?

I note I can set the random state in random forest classifier, but not in Complement Naive Bayes. I get error

TypeError: __init__() got an unexpected keyword argument 'random_state'

I assume this is because naive_bayes is computing the probability of each class?

from sklearn.naive_bayes import ComplementNB
model= ComplementNB(random_state=0)

Upvotes: -1

Views: 1389

Answers (2)

Simone
Simone

Reputation: 615

Three years on, sklearn has a page with information and code on reproducability. So you do not need to define your own function. It covers the two packages numpy and random, which are the underlying random generators. As amirhe said, this classifier has no option to define a random state (see documentation).

import numpy as np
import random

random_seed = 54 # Random Seed at file level

np.random.seed(random_seed) #set random seed for numpy
random.seed(random_seed) #set random seed for random

Upvotes: 0

amirhe
amirhe

Reputation: 2341

That's because the model doesn't do anything needed to be randomly initialized or generated. see the source code for more details.

Be in case you want to reset the random seed in that line from your own side, you can do something like this.

import numpy as np
import random

def set_random_seed(seed=0):
  np.random.seed(seed)
  random.seed(seed)

Then you can reset the random seed right before calling your model

from sklearn.naive_bayes import ComplementNB

set_random_seed(0)
model= ComplementNB()

Read more about the mechanism of Complement Naive Bayes.

Upvotes: 1

Related Questions