I have a dataset for which I'm using different forecasting methods (Ridge/Lasso Regression / Random Forests / AdaBoost / Gradient Boosting) to compare their performance.
For the full dataset, I achieve some good results overall in terms of the MAPE and the R^2.
But when I apply the regression to a specific subset of the dataset, the performance is much worse. I know it's impossible to tell the specific reasons for this without having access to the data but what are some general reasons that such a performance gap can exist?
Thanks in advance!