A Novel Chaotic Binary Butterfly Optimization Algorithm based Feature Selection Model for Classification of Autism Spectrum Disorder

Autism spectrum disorder (ASD) issues formidable challenges in early diagnosis and intervention, requiring efficient methods for identification and treatment. By utilizing machine learning, the risk of ASD can be accurately and promptly evaluated, thereby optimizing the analysis and expediting treatment access. However, accessing high dimensional data degrades the classifier performance. In this regard, feature selection is considered an important process that enhances the classifier results. In this paper, a chaotic binary butterfly optimization algorithm based feature selection and data classification (CBBOAFS-DC) technique is proposed. It involves, preprocessing and feature selection along with data classification. Besides, a binary variant of the chaotic BOA (CBOA) is presented to choose an optimal set of a features. In addition, the CBBOAFS-DC technique employs bacterial colony optimization with a stacked sparse auto-encoder (BCO-SSAE) model for data classification. This model makes use of the BCO algorithm to optimally adjust the 'weight' and 'bias' parameters of the SSAE model to improve classification accuracy. Experiments show that the proposed scheme offers better results than benchmarked methods.