add dynamic sensitivity for anomaly tests

[data-blade]

Is your feature request related to a problem? Please describe. anomaly tests require a set sensitivity. but a given sensitivity of e.g. 3 will 'look' like a real anomaly when plotting a volatile time series (high stdev), but no so much for flat time series (low stdev). i.e. the sensitivity a product person needs depends on the volatility of the metric, especially if they base their judgement purely on the visual impression of the plot. one sensitivity might yield to many false positives for them, and another too many false negatives. the additional config ignore_small_changes does not solve for this, unless specifically tailored for that metric.

Describe the solution you'd like instead of asking for a specific sensitivity, give the option to calculate it dynamically as the square root of the signal-to-noise ratio sqrt(median/stdev)

Describe alternatives you've considered setting different sensitivities for each metric, but cumbersome and hard-to-know before (and might still be the 'wrong' sensitivity for tomorrow's data)

Additional context from personal experience with various solutions (custom, expectations, elementary) on many time series this seems to yield the best stats for 'visual' false positives/negatives.

Would you be willing to contribute this feature? if time allows

[github-actions[bot]]

This issue is stale because it has been open for too long with no activity. If you would like the issue to remain open, please remove the stale label or leave a comment.