Genetic Algorithm Based Semi-feature Selection Method

Semi-supervised learning mechanism requires new feature selection methods to work on unlabeled samples. Traditional researches deal it with the help of ldquofilter-typerdquo semi-feature selection mechanism, which may not work well for classification tasks. Genetic algorithm is one of widely used ldquowrapper-typerdquo supervised feature selection methods. Here, we propose a novel genetic algorithm based semi-feature selection method. In essence, it uses unlabeled samples to extend the initial labeled training set with the help of classifiers, and with the feedback of classifiers, it can select more discriminative features for classification. Extensive experiments on publicly available datasets show that our proposed method outperforms both traditional supervised and state-of-the-art ldquofilter-typerdquo semi-feature selection algorithms.