Feature selection is a common and key problem in many classification and regression tasks. It can be viewed as a multiobjective optimisation problem, since, in the simplest case, it involves feature subset size minimisation and performance maximisation. This paper presents a multiobjective evolutionary approach for feature selection. A novel commonality-based crossover operator is introduced and placed in the multiobjective evolutionary setting. This specialised operator helps to preserve building blocks with promising performance. Selection bias reduction is achieved by resampling. We argue that this is a generic approach, which can be used in many modelling problems. It is applied to feature selection on different neural network architectures. Results from experiments with benchmarking data sets are given.

Genetic Algorithms are often employed for neural network feature selection. The efficiency of the search for a good subset of features, depends on the capability of the recombination operator to construct building blocks which perform well, based on existing genetic material. In this paper, a commonality-based crossover operator is employed, in a multiobjective evolutionary setting. The operator has two main characteristics: first, it exploits the concept that common schemata are more likely to form useful building blocks; second, the offspring produced are similar to their parents in terms of the subset size they encode. The performance of the novel operator is compared against that of uniform, 1 and 2-point crossover, in feature selection with probabilistic neural networks. 1