search

Brett-Kennedy / CountsOutlierDetector

Interpretable, intuitive outlier detector intended for categorical and numeric data.
MIT License
9 stars 3 forks source link

[question] time series #4

Closed lkaniak closed 1 month ago

lkaniak commented 1 month ago

Hello there,

I want to know if you ran any tests on time series and how it performs in such data. Does it provide a reasonable explanation about an anomaly? Have you seen the alibi-detect packages, and if so how this compares to them?

Thanks for your time, great post on TDS tho.

Brett-Kennedy commented 1 month ago

Yes, I cover time series data in my book, Outlier Detection in Python, https://www.manning.com/books/outlier-detection-in-python.

It's quite possible to do outlier detection with time series. Counts Outlier Detector is designed for tabular data, which is a bit different, though time series data can be converted to tabular format.

I have used Alibi-Detect and cover that in the book as well. In general, it works well, but all outlier detectors are a bit different, and all emphasize, and all miss, different types of anomalies.

Brett

On Fri, Jun 21, 2024 at 1:06 AM Lucas Kaniak @.***> wrote:

Hello there,

I want to know if you ran any tests on time series and how it performs in such data. Does it provide a reasonable explanation about an anomaly? Have you seen the alibi-detect packages, and if so how this compares to them?

Thanks for your time, great post on TDS tho.

— Reply to this email directly, view it on GitHub https://github.com/Brett-Kennedy/CountsOutlierDetector/issues/4, or unsubscribe https://github.com/notifications/unsubscribe-auth/AUHC56RDJVSCP3INE34AAVTZIOYEHAVCNFSM6AAAAABJVFTOSCVHI2DSMVQWIX3LMV43ASLTON2WKOZSGM3DKNZSGI2DANQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>