Predictive accuracy is repeatedly cited by data scientists as one of the most important demands in modern data mining algorithms and software. It stands right along side the importance of model-building speed, missing value handling, and memory efficiency. So, if it is so important, how do the experts TEST the accuracy of their models?
Software Advice (www.softwareadvice.com/bi/), a business intelligence research firm, recently interviewed top data mining experts, Karl Rexer, Dean Abbott and John Elder, to find out how they test the validity of their predictive models.
View the report below to see how the use: Lift charts and decile tables to compare performance against random results Target shuffling to determine validity of the results Bootstrap sampling to test the consistency of the model.
This post was originally published on January 29, 2014. You can read the full article here.