ML forecasting feature

Hi, I would like to know will odfe ML support forecasting in the future?

hi, @bpavani, can you help answer this question?

Hey @ftlj,

Yes- we do plan to support forecasting in the future. Currently we are working on trend analysis of historic data and we will be building on this to support forecasting. We would love to learn more about your use case if you could share.

Thanks,
Pavani

My use case is base on historical system performance metrics such as cpu or memory utilization to forecast the coming days or weeks figures. Maybe trend analysis with predictive function already fulfill my need.

Hi,are there and news about it ?
I need to di a forecasting about product that i have to buy, can i use opensearch ?

hi, @Oltre thanks for your interest. We do plan to support forecasting, and possibly that will be in this new ml-commons plugin https://github.com/opensearch-project/ml-commons.

We’d like to gather your requirements first to support you better. Can you elaborate your use case and expectation? Like just time-series forecasting and how long do you prefer to predict?

Hello, we are looking for an algorithm that will monitor our transaction flow. We have some peaks e.g. Hourly - at 11 to 13 and 16-18, at holidays, before Christmas, and wanted Machine Learning, that will predict these Holidays and change average flow at the season and that will not alert anomalies with serious severity when we’ve got some hourly peak. I tried Machine Learning 1.2.4 plugin and got alerted every day at the same time, but everything was fine. I need to control transaction flow, if we haven’t some downtime and if we don’t have more than 1% Declined payments in real time.

vnovotny98@, maybe you can try AD in the upcoming release (2.1) to see if things improve. Another thing is to increase shingle size (e.g., 12). Or you can wait for https://github.com/opensearch-project/anomaly-detection/issues/562.

@vnovotny98 Did you have a chance to review @kaituo’s proposal by chance?

@jdbright
Hello, we found an external company for this task.
They sent us variants of the target architecture that could work and we will test them, I will let you know when we come up with something.
We will test these. (it’s in Czech.)

For those who are tracking the Forecasting feature. I’ve opened up a Github meta issue to track the progress. We are eager to hear about use cases the community is interested in solving for.