A Joint Optimization Algorithm for Feature Selection and Classifier Design

As datasets continue to expand, the significance of feature selection in identifying influential features for classification becomes increasingly apparent. Meanwhile, the performance of a classifier has a great impact on the classification accuracy. As a simple and effective classifier, the extreme learning machine autoencoder (ELM-AE) has a wide range of applications. This paper proposes a joint optimization algorithm (JOA) for data classification, in which a hybrid genetic algorithm is employed to find an optimal feature subset and the optimal classifier. A binary-real encoding scheme is used to represent feature set of data, the activated state of hidden nodes and parameters of ELM-AE. The experimental results show that the proposed method can identify optimal feature subset and classifier parameters.