Any ideas on how to get started?
So, suppose you have some ratings on 10 different snowboards, each with
an 'appropriateness' score for some dimension of a snowboards attributes
(good for tricks, overall speed, degree of control, length, cost - I
don't know what's important, that's your domain knowledge, not mine). [1
model for the boards]
I assume you also want to have some questions that get at the notion of
how important each attribute is to a potential buyer. Likert scales seem
appropriate - these are the typical four or five point scales, such as:
Very Important, Important, No Opinion, Unimportant, Couldn't Care Less.
Four point scales force a choice on the plus or minus side, five point
scales allow a 'no opinion' in the middle. [another model for the your
When the user submits, score each board on each dimension, summing its
results, then recommend the board with the highest score (assuming more
important == higher value). I'd probably display the top three boards,
each with their score.
Now figuring a board's score is where the interesting stuff lies, IMO.
You probably want to start with a simple system for combining the board
appropriateness on a dimension with the user's rating of importance. An
easy approach up front is to just multiply the board appropriateness
score by the user's importance score. If a user rates something as very
important, and a board is supposed to be great for that, then that
should count highly into the overall score. If the user thinks an
attribute is important, but the board is rated as mediocre on that
attribute, then it should contribute less to the overall score. It will
probably take some tinkering and judgment to come up with decent scores.