search

business-science / anomalize

Tidy anomaly detection
https://business-science.github.io/anomalize/
338 stars 60 forks source link

Question: How to identify anomalous trends? #35

Open pedropenzuti opened 5 years ago

pedropenzuti commented 5 years ago

Hi folks.

First off, I'd like to give a big thank you to the maintainers and contributors of this package. It works great!

However, I do have a question: how would I go about identifying anomalous trends as opposed to transactions ?

Say I have 5 companies and I'm interested in identifying if, amongst those, there is one exhibiting an anomalous trend. Or if I'm selling a group of products, and I'm interested in identifying if any of those have anomalous trends compared to each other. How would I do that?

Again, thank you for the package.

Cheers

McCompexpert commented 5 years ago

Have a look at this one: https://github.com/robjhyndman/anomalous-acm and this paper https://robjhyndman.com/papers/icdm2015.pdf

pedropenzuti commented 5 years ago

Will do. Thank you for the quick reply!

mdancho84 commented 5 years ago

It sounds like you may be interested in change point analysis, which is the process of detecting changes int trends. Check out the changepoint R Package. https://r-datameister.blogspot.com/2013/08/changepoint-analysis-of-time-series.html