What's the difference between clusters and profiles

[Rocko]

There is little information available about clusters, also Analytics Builder does not provide a GUI to create them, you can only go via API.

To my understanding, clusters group the features with respect to the goal. What I don't understand is what is the difference to profiles then, which do the same thing.

So what are the conceptual differences, and when do I use which?

Also, when you look at the result of a cluster job, in one cluster, you only get only a list of features with their distinctiveness. To me, this doesn't describe the cluster very well, I would have expected to get something like a "center of gravity" for the cluster to be able to say where in the n-dimensional feature space the cluster located. How can I get there?

It'd be great if someone could shed some light on these two issues.

[mchehaibi · Accepted solution]

Hi Roman,

In ThingWorx Analytics the Profiles are representative groups of examples that have over or under performed against the goal historically:

• This Uses historical (or back-looking) data to establish the performance of Profiles.
• Similar to signals in that it is like a magnet on data patterns, but reveals the combinations of factors that demonstrate higher or lower performance of a Topic.
• ThingWorx Analytics uses a search algorithm that builds and finds groups that meet the required conditions for a profile
• They must be large enough in size
• They must over or under perform by a defined threshold

Whereas the Clusters are groups of objects within the Dataset the behave in in a similar way or are close to each other when compared to the rest of the object in the Dataset. This similarity in behavior or proximity in distance is measured against the goal and also against other features in the Dataset (not only the goal unlike profiles). It is basically an approach of Explanatory Data Analytics.

I hope this answers your question

Best Regards,

Amine