I have a question regarding the creation of structures and models and how to integrate them properly into our service platform.
I'm currently trying to create a reporting function that will also allow some predictive visualizations. Basically a couple of values along bars will be displayed and those will be equipped with sliders to serve as parameters that are inputs to another value, e.g.
Price / User_Rating / Distance ---> Page_Ranking
So I will throw all of these into an algorithm and try to predict the page_rank.
However, these reportings need to be tailored towards specific peer groups which will differ in one or more dimensions (e.g. the services you want to examine, the region you're in...) and the users will be able to specify the peer groups themselves.
I assume that, depending on the way you slice this data mining problem, the relevance of the individual parameters will differ. So my question is, is it possible to get a mathematical function (some kind of equation) out of the patterns that the model discovers and then I can just do some kind of WHERE statement to filter it, or will I have to manually create models for every possible combination of peer group parameters?
Thanks in advance,
- Edited by mmmariuss Monday, May 20, 2013 11:26 AM
Which algorithm are you planning to use?
Tatyana Yakushev [PredixionSoftware.com]
Download Predixion Insight 3.0 - World class predictive platform for big data
I just noticed that there's no such thing as multivariate regression. I'd like to use something simple that can take a few continuous input variables and predict a continuous output variable. Is there such a thing within the SSDM suite?
However, this does not really pertain to my above question: Will I have to create new models for every specific type of reporting? Or is it possible to have one model that is parameterized in the right way (like a singleton query)?
- Edited by mmmariuss Tuesday, May 21, 2013 4:25 AM