What do the statistics mean?

Mean Squared Error: The average squared difference between actual and predicted values
|    - The higher the value, the less accurate the model is.
R2 (R-Squared): How well the regression tree fits the data.
|    - The closer to 1, the better
|    - Despite my models having a high R2, this does not directly equate to it's accuracy. There are more factors that go into SR and the SR economy will always change.

The first two models...
I only had the Random Forest model, which was good for what it's meant for: interpolation, guessing a value within the training data. However, going outside the SR range for my data resulted in no extrapolation, guessing a value outside of the training data.

This meant that the model was really good at returning accurate results for the 2K-17K range, but not good for anything outside of that. The model would never return values under 2K or above 17K. Therefore, I decided to introduce a model that can "guess" values outside the training data's range.

Rather than replace the model entirely, I decided to have two models instead, switching to the alternate model which supports extrapolation when needed.

...and the rest?
The rest are experimental models, essentially made out of curiousity. Some of these are... less practical than others
|    - Gradient Boosting
|    |    - Similar to the Random Forest, but rather than each "tree" being independent, the trees learn from each other and improve based off of that.

Font used: Space Mono
Created with Microsoft FrontPage 2003

Home || About