search

Marcnuth / AnomalyDetection

Twitter's Anomaly Detection in Pure Python
Apache License 2.0
304 stars 76 forks source link

possible to support sec level of granularity? #26

Open hokiegeek2 opened 5 years ago

hokiegeek2 commented 5 years ago

Hi, was wondering if it is possible update this algorithm for seconds level of granularity?

mariojst commented 5 years ago

Hey @hokiegeek2

In my case, when wanting to evaluate two levels of granularity, in fact we still only require a single analysis to be made. An example to be clearer:

Having a dataset of Dimension_1 | Dimension_2 | KPI X1 | Y1 |10 ...

What we want to understand is if on combination X1+Y1 there is an outlier. Therefore transforming the dataframe to only have: Dimensions | KPI X1,Y1 | 10

performs the exact same intent of the analysis, while passing only a single level of granularity.

Sorry for typos or formatting of the answer, hope it was helpful :)