John Elder’s presentations on common data mining mistakes are a must-see if you have any experience or plans in the data mining arena. One example I recall in particular – a neural net that performed brilliantly in distinguishing the presence or absence of tanks in photographs. The only problem? The tank photos were taken on sunny days, the non-tank photos on “puffy-cloud” days. The algorithm was focusing on the clouds. John has an online course with many such examples, and much more to share, based on years of experience with industry, at statistics.com – Oct. 28 – Nov. 14. See https://www.statistics.com/dmmistakes .
