anomalydetection.mad() function

anomalydetection.mad() uses the median absolute deviation (MAD) algorithm to detect anomalies in a data set.

Input data requires _time and _value columns. Output data is grouped by _time and includes the following columns of interest:

• _value: difference between of the original _value from the computed MAD divided by the median difference.
• MAD: median absolute deviation of the group.
• level: anomaly indicator set to either anomaly or normal.

Function type signature

(<-table: stream[B], ?threshold: A) => stream[{C with level: string, _value_diff_med: D, _value_diff: D, _value: D}] where A: Comparable + Equatable, B: Record, D: Comparable + Divisible + Equatable

{{% caption %}} For more information, see Function type signatures. {{% /caption %}}

Parameters

threshold

Deviation threshold for anomalies.

table

Input data. Default is piped-forward data (<-).

Examples

Use the MAD algorithm to detect anomalies

import "contrib/anaisdg/anomalydetection"
import "sampledata"

sampledata.float()
    |> anomalydetection.mad(threshold: 1.0)

{{< expand-wrapper >}} {{% expand "View example input and output" %}}

Input data

Output data

{{% /expand %}} {{< /expand-wrapper >}}