ML Population Analysis

[saraKM (Sara)]

Hi all,

I have a population model created as per recipes with configuration as follow:

"high_info_content(destination.subdomain) over destination.domain"

I'm trying to understand how the model is working for this specific job. As I understand from temporal/population analysis article, this will create a typical entropy value based on the subdomain analysis of all destination domain values seen in that bucket. So just wondering if this typical value is only based on the subdomains seen in that bucket or considers the history of those values? also for each destination domain X in the bucket (analyzing the subdomain information for domain X) there will be a new model created for all other destinations excluding X as comparison baseline?

Thanks,
Sara

[richcollier]

The model incorporates all entities (in this case, domains) seen in each bucket, for all time that the job has been running. In population analysis, only this "collective" model of all entities is made (and not one per entity). But, every entity within the bucket is compared against the collective model. This is more efficient for very high cardinality entity sets.