Magazine article Information Today

Data Mining: Life after Report Generators: Libraries Use This Decision-Support Technique to Chart a Future Course

Magazine article Information Today

Data Mining: Life after Report Generators: Libraries Use This Decision-Support Technique to Chart a Future Course

Article excerpt

Data mining, the term itself, evokes strong images in my mind's eye. I envision little characters resembling those comical fellows pictured in Mad Magazine's "Spy vs. Spy." In my imagination they've replaced their wide-brimmed hats with spelunkers' headgear (headlights and all), and they're lowering themselves on cable strands to the inner depths of my disk drives. Once there, these guys begin frenetically digging up chunks of data resembling precious gems. When they have buckets full of gems, they eagerly climb back up through my computer toting the gems behind them. Then what?

What Value, These Gems?

I wonder what they're going to do with the gems. Are the gems more precious than the tidbits we used to retrieve with yesterdays' report generators? To say data mining gems are more precious than report generator gems is too simple. But, data mining algorithms allow for interpretation and prediction analysis based on information in our databases, the likes of which we haven't seen before. This type of analysis moves beyond Structured Query Language (SQL) reports into the realm of more machine intelligence.

Data mining techniques don't stand on their own. They are pan and parcel of new system developments and ways of looking at the challenges of working with large databases. The bigger picture includes data warehousing, target data selection, model selection, evaluation, interpretation, OLAP (Online Analytical Processing) servers, and the like.

Making the transition from older systems to accommodate these fascinating new possibilities is not just a matter of porting an existing database. There are important data-migration and upgrade issues to be undertaken before moving to a more sophisticated new generation system. Not only do we have the elements (fields) in the database to consider, but for sophisticated data retrieval and analysis, we will have to be aware of having programs to capture metadata. Metadata is data about the data fields. Metadata lets the system know everything about how data was obtained, how it was defined, what attributes are contained, and so forth. Also, data consistency is extremely important to achieving optimum results.

Perhaps something this sophisticated is not what your library needs, and not something you'll want to invest in, but if you're working with very large databases or with specialty databases, you'll want to consider making the investment in a practical decision-support system. As our databases grow and constantly change, it's become almost impossible to spot trends and changing patterns manually, not to mention quickly enough to make a difference in optimizing collection development, or in providing up-to-the-minute or new information services.

Data Mining Spots Changing Trends in User Behavior

In what ways would this type of decision support information be of use to your library?

One way to use data mining in decision support is to better understand patron behavior, be it the CEO, a student, or an occasional borrower. Behavior patterns are not only concerned with borrowing, but if your system uses ID or account numbers for remote login, tracking, or for access to certain databases, then you have the possibility of constructing a complete picture of how your resources and services are used. What sets data mining apart from report generation is its ability to take this information and go beyond stating what was and into the realm of predicting what could be. There are also 3-D visualization tools that can be used to look at results from all angles, as automatically predicted by the program. Once you're able to predict patterns based on previous behavior, you have the opportunity to meet future demands more successfully.

Not only will you be able to refine collection development to target more specific needs, but you'll be able to design library programs based on known behaviors and affinities. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.