Like planning, data mining (AKA KDD (Knowledge Discovery and Data Mining)) is kind of leaning over the edge of facts to select from a set of more or less risky possibilities. So besides being something that planners might find useful, it's also a field that planners might find interesting, in part because KDD's efforts to automate the process of finding the most valuable among the huge number of patterns discernible in any large database has led them to contemplate the nature of interestingness.
There is no standard definition of interestingness among the miners. That would go against their gritty and perverse nature. But there is wide acknowledgment that interestingness is central to their task (“Data mining can be described as the process of finding interesting patterns in large databases”) and that it is usefully decomposed into a number of sub-criteria:
Interestingness differentiates between the "valid, novel, potentially useful and ultimately understandable" mined association rules and those that are not--differentiating the interesting patterns from those that are not interesting. Thus, determining what is interesting, or interestingness, is a critical part of the KDD process.
Different authors produce different lists of criteria...
...but in general, the miners' blueprint of interestingness overlaps to a surprising degree with intuition and my more scribbly formulations. Novelty, validity, surprisingness, range, peculiarity(!) and utility are all there, each one intriguing if not entirely self-explanatory (I see that as my job), and the "ultimately understandable" criterion is consistent with the observation that delayed comprehension can be more valuable than immediate understanding. At least, it should make us reconsider our definition of comprehension.
Insights into interestingness are everywhere, but data mining offers an especially rich, data-driven yet thoughtful lens through which to view and address our own interestingness management issues. What makes "Just Do It" such a concise, surprising, useful idea to so many people? What contextual knowledge ("domain expertise" in minerspeak) is required for a campaign idea to be novel rather than just weird? "Novel, surprising and peculiar, yet valid, useful and ultimately understandable" could describe a leverageable pattern of purchase behaviors mined from the slag of last week's retail numbers. It could also be a critic's response to Ornette Coleman's Free Jazz or a viewer's take on a chocolate ad featuring a musical gorilla. Each of these things is worth minding and potentially valuable because first of all, they are interesting.
Comments