Anomaly detection

mskh

Hi,
I use histogram-based outlier score for anomaly detection but it dose not consider the dependencies between features. Which anomaly detection algorithm should i use to consider the dependencies between features?
Thanks

Telcontar120

That is the nature of HBOS. I am not entirely what you mean by "dependencies" in this instance, but you should consider one of the distance or density-based outlier approaches such as k-nn Global Anomaly Score or Local Outlier Factor (LOF) which are inherently multivariate in nature. Be sure to normalize your data first though since they are scale-sensitive!