Non-standard crossover for a standard representation--commonality-based feature subset selection
Stephen Y. Chen, César Guerra-Salcedo, Stephen F. Smith · 1999
The Commonality-Based Crossover Framework has been presented as a general model for designing problem specific operators. Following this model, the Common Features/Random Sample Climbing operator has been developed for feature subset selection--a binary string optimization problem. Although this problem should be an ideal application for genetic algorithms with standard crossover operators, experiments show that the new operator can find better feature subsets for classifier training. 1 INTRODUCTION A classification system is used to predict the decision class of an object based on its features. When training a classifier, it is beneficial to use only the features relevant to prediction accuracy, and to ignore the irrelevant features [Koh95]. The benefit arises from an increase in the "signalto -noise ratio" of the data, and a reduction in the time required to train the classifier. Thus, the objective of feature subset selection is to identify the (most) relevant features. Feature sub...