Spatial Data Mining and Geographic Knowledge Discovery

Unlike some other fields in GIScience, advances in spatial data mining and geographic knowledge discovery are not only needed, but time sensitive. The rate at which data is collected and produced is accelerating with little end in sight. This is not due only to the number of observations, but the number of times an observation is made. Montreal’s public bus system, for instance, was in the dark ages until only a year or two ago. Now data is constantly collected from bus-mounted GPS units [Amyot]. At this rate we GISystems could drown in the surge of oncoming data. That is not to say that excess data is a bad thing. In a world in which one can must choose between too much and too little data, too much, I think, wins out. That doesn’t mean an excess of input is not a double edged sword.

Algorithms, data structures, and hardware limitations constrain the future of the science and must be improved upon. On the note of a double edged sword, however, it is my only guess that as these factors are improved, the incoming stream of data will only increase as well. What worries me is statements, like the one made by Guo and Mennis, “more recent research efforts have sought to develop approaches to find approximate solutions for SAR so that it can process very large data sets.” I understand that many times projects may have deadlines, researchers may have other places to be or feel obliged to not hog all the computing power. At the same time, the benefits to computing algorithms using complete likely outweigh the computing costs. Then again, the largest data set I have ever created on my own was an excel spreadsheet no bigger than 1 megabyte.

AMac

Comments are closed.