'random_state' parameter for Kmodes

[nicodv]

'random_state' parameter for Kmodes

[nicodv]

There now is a random_state argument available.

Old answer:
Yes there is. Set the random seed for numpy: np.random.seed(42)

[Nishanth14]

Thanks! it worked.

[supreetkt]

Hi, I have a similar issue. I'm able to set a random_state parameter, but it isn't working as a seed. The results change with each run anyway.

[nicodv]

@supreetkt , could you present a minimal reproducible example?

[supreetkt]

Okay, I think I was wrong. I'm not able to set the seed at all. Currently, my predictions for kmodes are changing with each run.
Can you tell me how to set a seed for k_modes here? I tried:

kmodes_seed = 123
km = KModes(n_clusters=k, init='Huang', n_init=5, random_state=kmodes_seed)
model = km.fit(np_array)
predictions_lst = km.labels_.tolist()

and it isn't allowing me.

[supreetkt]

Okay, I figured it out. I had installed kmodes with pip install and somehow it still contained the previous version where random_state was not available as a parameter. I updated the file and it is working now.

[nicodv]

OK, great. Yes, that feature will be in the next release.