I ran a Polynomial multiple regression model on a dataset with just 98 points. when I ran the model on different subsets of data (training and testing). It gives me r2 value ranging from negative to positive.I know I have fewer number of data points. I want to ask if I can stick to the run with the highest positive r2 value. Also, how do I save that model if I need to run it later on? Because when I closed the r window and ran the code again, I didn’t get the same r2 value and results but a different highest positive r2.

I am working in r, and I tried other different models as well, like random forest and SVM. But I find PMLR best. But I am getting inconsistent r2 values.