Hi Monika,
I find this K20 lab CCW1281: Improve your agents' efficiency with Machine Learning based Similarity Solutions helpful in understanding the difference between training frequency and update frequency. Around 28:00 mark in the video, Andrew Wong explained what the Update Frequency means. As per the lab guide, it's "how often you want to refresh (on a rolling basis) the data set."
With each re-training, you start training your solution fresh based on the filter criteria, and the previous solution is discarded. With update, the system checks for any newly created/updated records since the last update and add them to the existing clusters.
For both frequencies, the "discard all previous cluster results" as per your screenshot above is confusing to me too. I would think that with update, previous cluster results are not discarded.
From the Machine Learning Success Playbook, the descriptions for the frequencies make more sense:
"Refresh and Recluster Frequency – In these fields, select how often you want the system to group new and updated records into clusters (aka update frequency) and how often you want the system to discard results and recreate clusters from the beginning (aka training frequency)."
Thanks,
Jenny
