I'm looking for a CHAID (Chi-square automatic interaction detection, decision-tree algorithm) implementation in Python that supports setting a parameter that is equivalent to "random_state" so that I can perform a reliable hyperparameter optimization. "Random_state" is a parameter in scikit-learn which can be set to avoid random influences on the calculated models, which I'd like to use to make sure differences between models are caused by different sets of parameters. Additionally, I need the package to be able to produce probability estimates per case and an overview of the model (graphically or as text), but I suppose those features are more common.
I'm not exactly confident that I can take any CHAID package out there and implement a random_state equivalent parameter myself, which is why I wonder whether there are any packages available that support this out of the box. Can anyone point me into the right direction? Any help is much appreciated!