Why do unusually short isolation-tree paths suggest anomalies?
Classic MLLevel: IntermediateRandom isolation depth
Why do unusually short isolation-tree paths suggest anomalies?
Hints
- Compare isolating a remote point with separating a dense cluster.
- The statistic is a path length.
Answer
Isolated observations can be separated from others by few random partitions. The forest aggregates path lengths rather than fitting class labels or assuming a Gaussian density.
Source and licence: scikit-learn 1.7.2 — outlier detection · BSD-3-Clause · answer: Quantoros editorial