Government report: Data mining doesn't work well

I don't think there are enough cases of terrorism to be able to apply data mining to this problem. There just isn't enough statistics and every case is different.

I was at Oracle at the peak of the data warehouse/data mining hype cycle. Every presentation cited the correlation found between sales of beer and nappies (diapers). The only problem is, it probably never happened.

http://web.onetel.net.uk/~hibou/Beer%20and%20Nappies.html