Model Training & Adaptation
Gradient boosting
Combining prediction systems added one after another, with each helping reduce the remaining errors.
Example
Successive small trees refine a demand prediction.
Why people use it
It improves a combined predictor through a series of smaller corrections.
What you'll hear
“Each new part tries to improve the earlier prediction.”
What this means for you
Monitor overfitting and compare performance on relevant unseen cases.
Can you control it?
Developer-only
The people building or running the AI choose this setup. An everyday user generally needs their help to change how this part works.
Common questions
- Does the method always use question-based decision trees?
- No. Trees are common, but other prediction components can also be used.
- Does adding more parts always improve new predictions?
- No. Too many can fit quirks in the teaching examples rather than useful general patterns.
- Can the small parts be simple trees?
- Yes. Small decision trees are commonly used, though the broader idea is not limited to that choice.