Anomaly Detection
Detect anomalies and outliers using a Digital Twin.
Flags rows a digital twin did not expect, and says which variable makes them surprising.
Because the expectation comes from the twin rather than from a threshold, a row can be anomalous at values that are individually ordinary — it is the combination the model rules out. Note that the anomalies port carries every scanned case and not only the flagged ones.
At a glance
Module type
AnomalySim
Category
Intelligence
Default tier
durable
Runs on
reactive, durable
Starts a flow
no
Holds a body
no
Keeps state between runs
no
Available in
every application type
Parameters
digitalTwinId
string · required · defaults to ""
The twin to run against.
digitalTwinName
string · optional
Its name when it was chosen, for the node's label.
digitalTwinVersionIds
string[] · optional
Pin the run to specific twin versions; empty means the latest.
scenario
Record<string, unknown> · optional
The saved scenario. Anything arriving on the scenario port is merged over this.
scenarioDraft
Record<string, unknown> · optional
What the inspector is editing, before it is applied to scenario.
scenarioValidationErrors
string[] · optional
What was wrong with the draft the last time it was checked.
entityColumns
object · optional · defaults to {"id":null,"target":null,"constraints":null,"levers":null}
Which columns of the entities frame carry what: id the environment id, and target, constraints and levers the optional per-row overrides. Any may be null.
Inputs
on
event
no
Implicit on every module that does not itself start a flow: wire an event here to make this module the start of what that event runs.
scenario
json
no
Sensitivity and the window to score, per run.
entities
dataframe
no
One row per environment to scan; entityColumns maps its columns.
baseline
dataframe
yes
The rows to score. Required: there is nothing to flag without them.
Outputs
anomalies
dataframe
dataframe, signal
One row per scanned case, flagged or not — is_anomalous is a column, and nothing filters on it. Also entity, score and flagged_variables. Filter it yourself.
summary
json
json, scalar, signal
The run's identity, not its findings: {runId, status, outputs}, where outputs is the simulator's raw per-version result. The numbers a reader wants are in results.
Last updated