A self-adaptive multi-objective feature selection approach for classification problems

In classification tasks, feature selection (FS) can reduce the data dimensionality and may also improve accuracy, both of which are commonly treated as two objectives in FS problems. Many meta-heuristic algorithms have been applied to solve problems they perform satisfactorily when problem is relatively simple. However, once datasets grows, their performance drops dramatically. This paper proposes a self-adaptive multi-objective genetic algorithm (SaMOGA) for FS, designed maintain high even grows. The main concept SaMOGA lies dynamic five different crossover operators evolution process by applying mechanism. Meanwhile, search stagnation detection mechanism proposed prevent premature convergence. experiments, we compare with on sixteen datasets. According experimental results, yields set well converged distributed solutions most sets, indicating that guarantee while removing many features, advantage over its counterparts more obvious