Description
Random Forest is an ensemble of decision trees trained on bootstrapped subsets with random feature splits, aggregated by majority vote (classification) or averaging (regression). It is the canonical strong baseline for hand-crafted CSI feature pipelines because it is non-parametric, requires almost no tuning, and gives feature-importance scores for free.
When it's used
- Strong non-DL baseline for occupancy / HAR / gesture
- Feature-importance analysis on subcarrier-statistics inputs
- Sensor-fusion classifiers combining CSI with other modalities
- Quick-and-dirty deployment on edge hardware
Limitations
- Underperforms deep models on raw / weakly-engineered CSI
- Memory footprint grows with tree count
- Cannot model time-series structure without windowing