How can we feed the dev/test data as historical data to ML job for production

Here i have a case like where for every small change in my data is treating it as anomaly in production server. so i want to feed the last one year data to my ML job in Dev server and the machine learns the data and categorise the anomalies. Now i want to transfer this to prod server, so that i can get the accurate results in detecting anomalies, otherwise it treats as anomaly for every minute changes happen.

Any suggestions helpful !!!

We understand that moving a job (and its assets like its model) is a desired feature - we're working on that.

However, with that said, your description implies that the data in dev/test is vastly different than the data that is in production. Keep in mind that the ML model is always "learning" and even if you were able to train in dev/test and move to production, the behavior or the data in production would ultimately still be learned and the model adapted.

So, my next obvious question is why you feel you aren't getting good results in production. Are you not letting the job learn over a sufficient amount of time (perhaps more than 3 weeks?)

Can you be more specific about the use case, the data, and the ML job config?

we did not deploy the code in production still. But while i am testing in preprod/test env, i observed like for minimal changes in data also treating as anomaly.
yes, its a all about 10 days data so far in test env.

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.