I compare regression tree and ols in terms of out of sample prediction. I realized that the mse values changed when i change the seeds before getting train and test set. Sometimes ols is better sometime regression tree. After i realized that i did a for loop for 100 different seeds and computetd 100 models and did 100 predictions and averaged the mse values at the end. Now regression tree is better.
Now i ask myself how can i get a good model out of this approach and can i say now that on average the regression tree outperforms the ols regression in terms of prediction?