Random Forest - Low bias and Low variance
- Random Forest works on selecting only a subset of features rather than selecting all features.
- These subsets of features might be less correlated to the target variables which in turn make models under-fit over the data.
- The features are selected randomly and this process happens over large iterations like 100, or 1000, or maybe 10,000.
- That's why some models get under-fit while some get over-fit.
- But at the end, these are combined to cancel out the effect of over-fitting and under-fitting.