Cluster Window (CW)

YAML type: cw | Category: Adaptive

Compares averages of consecutive sliding windows to detect level shifts.

Description

CW flags when the mean of one window differs from the next beyond eps, with enough points (minPoints).

How it works

Cluster Window (CW) illustration

Cluster Window compares the average of one sliding window to the next. When the absolute difference of those averages exceeds eps and both windows have enough points (minPoints), it reports a level-shift anomaly.

It targets step changes in the mean level rather than single-day spikes. Larger windowSizeDays smooths noise; larger eps reduces sensitivity.

Setup

Increase windowSizeDays for smoother shifts; raise eps to reduce sensitivity.

Configure detectors on the data view Analysis tab or in YAML. Validate with Data Preview before enabling production schedules.

Settings

Parameter	Default	Description
`windowSizeDays`	`7`	Sliding window length
`eps`	`1.0`	Minimum average difference
`minPoints`	`2`	Minimum points per window

Shared series filters

Most detectors also accept optional series filters in params:

Parameter	Purpose
`maxGapAmount`	Skip series when gap amount is too high (`null` = no limit). Catalog create-default is often `5`.
`minValue`	Skip series whose average value is below this (`null` = no limit). Catalog create-default is often `1000`.
`maxValue`	Skip series whose average value is above this (optional; omit/`null` = no limit)
`useSpecialDates`	When `true`, suppress anomalies that overlap special dates

When a filter skips a series, Data Preview shows the reason.

Configuration example

analysis:
  detectors:
    - type: cw
      params:
        windowSizeDays: 7
        eps: 1.0
        minPoints: 2

Using the detector

Interpret as a step/level-change detector rather than a spike detector.

Use cases

After pricing or catalog changes
Detecting new steady states post-incident
Gradual regime steps that spikes miss

Specifics

Related: changepoint, sustain.