I used Knime to make a predictive classification model using different tree-based models like RF, XGBoost and … With RF, every time that I run the model, I get different results which makes sense as I didn’t used a static random seed. However, with XGBoost (Tree Ensemble Learner node), every time I run the model, I still get same output with exactly same probabilities. I’m not sure what could be the issue as everything looks fine with RF but not with XGBoost. Has anybody experienced a same issue with XGBoost nodes/workflows?
I should also mention that I need to run the model several times to get the probability for each prediction to calculate uncertainty.