Andromeda Yelton

Across Divided Networks

data mining for fun and…

November 10th, 2009 · 1 Comment · Uncategorized

That slideset yesterday was funny, so I’ve RSSed the guy’s blog. Liked this recent post about data-mining your circ records. His university now has a recommender system (both “people who liked this book also liked” and “people in this course of study tend to like”) and a course-of-study-specific search functionality (nursing and law students want different books when they search for “ethics”). Turns out the recommender service is very popular and noticeably increases how much of their collection circulates (which my little ROI neurons like). Also provides suggestions for refining large searches based on search data. And keep an eye out for the very clever acronym which will warm your heart if you, like me, were online in the early ’90s.

Tags: ······

One Comment so far ↓

  • Newt Sherwin

    I would totally go for a “recommendations” feature at my local library. The “course of study” feature sounds a little less useful in that context, although if the software were flexible enough to instead use that for “career/interests” or the like and let users set their own, or simply have some way for tracking and learning from which search results have been most interesting for a particular user, I could see that having some utility. I wonder what my dad would think of this; I should forward the link to him. (Also, I would be very amused to see how a computer would label me, based on my library searches.)

    On a very different note, Chris works in data warehousing, and his company has recently branched out from that into data mining. I should likely point him in the direction of this post as well. :) But, you know, if you ever find yourself working in a context in which actual data mining people would be helpful, we probably know someone who could be helpful. He and I were discussing a while back the relationship between libraries and computers. Initially, like back when we were kids, libraries were looking for the best the computers could do, because they’d been trying to efficiently keep track of who had what book and when it was due, etc., for ever and now finally there was a machine that was close to being able to do it for them! Now, business applications have generally left them in the dust — computers got beyond being able to do what librarians had always been aiming for, and librarians have been slow to adjust their expectations to utilize the new features they could be reaching for. Likely, this is partly a matter of librarians not being the most tech-savvy people in the world — it’s not their major, after all — and of library systems being generally cash-poor, and nifty new features not really helping a library’s bottom line all that much. In the business world, economic pressures are pushing them to figure out how to better utilize the data they have (or could get), while most of those pressures don’t really apply — or apply much less directly — in a public library setting.

    Newt