Random cut forest anomaly detection on multi variant time series data

Question

I have sensor data coming from equipment with times series along with many attributes,

I have used RCF algorithm to detect anomalies. Now the challenge is,how to to convince the end user whether it is really anomaly or not. Just want to know which attribute is contributing to anomaly.

Is there any best way to convince end user whether it is really anomaly or not.

Answer 1

The simplest way to run the RCF model and to get the explanation for the anomaly is to use the version of RCF in Kinesis Analytics (KA). Here is a link to the documentation of how to run from the KA documentations: https://docs.aws.amazon.com/kinesisanalytics/latest/sqlref/sqlrf-random-cut-forest-with-explanation.html

Kinesis is taking care both for the training of the model, the inference after the initial training and for the attribution and explanation of the variables.