Citekey | MarteauEtAl2017Hybrid |
Source Code | https://github.com/pfmarteau/HIF |
Learning type | supervised |
Input dimensionality | multivariate |
Marteau, Pierre-François, Saeid Soheily-Khah, and Nicolas Béchet. "Hybrid Isolation Forest-Application to Intrusion Detection." arXiv preprint arXiv:1705.03800 (2017).
- ntrees: Number of trees in the HIF
- sample_size: Number of samples each tree receives from the input dataset
Because each tree is created by splitting along a random dimension at a random point in the data, adding more trees or increasing the sample size can help improving the confidence in the anomaly scores at the cost of performance.