Hello Mathworks community,
I am working on a script that I am translating from MATLAB to Python (Jupyter Notebooks). After reading in and prepocessing data inputs, I compute my X and Y data arrays that I want to feed into the train_test_split function. Then, I train my model (Random Forest regressor) using the training data and check how capabel my model is by checking predictions against test values. In MATLAB, I get correlation values of around 0.65-0.7, which is pretty good considering the data I am using. In Python, the R values are around 0.4-0.5.
The code is 1 to 1 translated (to the best of my abilities). But I am able to replicate the exact same X and Y input arrays from the original MATLAB code in Python, which tells me that my translation is correct so far.
Now my question/issue: Why do I get far worse correlation values using my Python model? It tells me that the Random Forest python equivalent is learning and predicting much worse even though I am technically using the same ML model and feed it the exact same data. Are there fundamental differences between the two platforms that I am not aware of? Or is it normal that different programming languages process ML models differently?
Thank you for your insight!
