Random Forest provides different results even if random_state is setted to a number

[romcenci]

Hi!

I'm using Random Forest Classifier to train an image segmentation model, but even setting the random_state variable, which according to the manual should give deterministic results, I get stochastic results that change every time I use the model to train and predict the same data set. In addition to random_state, I'm using the bootstrap flag and oob_score set to True. Is there any way to get deterministic results each time I run this model?

I also found some good discussions on these two forums:

• https://stats.stackexchange.com/questions/399819/what-makes-a-random-forest-random-besides-bootstrapping-and-random-sampling-of-f
• https://stats.stackexchange.com/questions/415416/how-does-a-decision-tree-model-choose-thresholds-in-scikit-learn

[glemaitre]

I think this is something that we are aware: #12188 (comment)

This is coming from tie-breaking.

[hrak99]

Use top of your code for producing same results,

np.random.seed(1111)

[hrak99]

You can choose any number like 42 instead of 1111.