Thinking about this more. Alternatively does it also make sense to build recommender based on these column attributes /values. So going back to below usecase when user picks a single item from the search results then find recommendations based on the values (for example in first case values of risk levels of a selected item)?
Or this is a better candidate of classifications? Again can you confirm that clustering isn't suitable for such usecases? Thanks Sent from Yahoo! Mail on Android
