Gene Selection Using Multi-objective Genetic Algorithm Integrating Cellular Automata and Rough Set Theory

摘要

Feature selection is one of the most key problems in the field of machine learning and data mining. It can be done in mainly two different ways, namely, filter approach and wrapper approach. Filter approach is independent of underlying classifier logic and relatively less costly than the wrapper approach which is classifier dependent. Many researchers have applied Genetic algorithm (GA) as wrapper approach for feature selection. In the paper, a novel feature selection method is proposed based on the multi-objective genetic algorithm which is applied on population generated by non-linear uniform hybrid cellular automata. The fitness functions are defined one using set lower bound approximation of rough set theory and the other using Kullbak-Leibler divergence method. A comparative study between proposed method and some leading feature selection methods are given using some popular microarray cancer dataset to demonstrate the effectiveness of the method.