Gisele L. Pappa, Alex Freitas: Automating the design of data mining algorithms, an evolutionary computation approach

摘要

The book proposes using grammar based genetic programming (GP) to evolve innovative data mining algorithms for classification tasks. These algorithms are then extensively tested on a number of data sets. What sets this approach apart from earlier methods is that previous attempts simply tune a numerical parameter, whereas this method actually generates new algorithms in the form of if-then-else statements. In other words, genetic programming is used to automate the design of rule induction algorithms, as opposed to manually designing them which is the conventional approach. Pappa and Freitas concentrate on easy-to-interpret classification rules of the form of "if (conditions) then (class)" as we (humans) are more likely to place confidence in human interpretable rules, rather than black box approaches such as artificial neural networks. Thus the book is about decision support rather than automated decision making.