Feature subset selection via multi-objective genetic algorithm
H.C. Lac, Deborah Stacey · Proceedings. 2005 IEEE International Joint Conference on Neural Networks, 2005. · 2006
Feature (attribute) selection is an important preprocessing step in data mining because most real-world databases were collected for purposes other than data mining. Consequently, many attributes in these databases are irrelevant for the data mining task. Moreover, some data sets have a curse-of-dimensionality problem and eliminating irrelevant attributes allows more accurate models to be built. In this thesis, a new system is proposed for the feature subset selection problem that is based on the wrapper approach. Specifically, this system works by wrapping our proposed multi-objective genetic algorithm around a regular backpropagation neural network.