Importance of 2 datasets

I am pretty new to machine learning. With respect to the problem I modeled my classifier on the basis of 2 hour interval data set. But I am not able to minimize my rmse anymore. What should I do next to improve on this. Is it something related to the classifier or something related to the existence of 2 data sets.
I am using regression forests right now.
Any kind of help would be appreciated.

Thank you.

Did you make 2 separate models for the two different cities?

Because the cities are far from each other, our experience is that 2 separate models work better.
See also the Benchmark submission in Python.