Novel Rider Hawk algorithm for feature selection in diagnosis of ASD using fMRI images

Abstract

The autism spectrum disorder being a disease that affects the children and has a no proper medicinal cure till date the diagnosis and care is the only way to improve such children. The disease has a spectrum of disorders and therefore identifying the affected group is a challenging task. In this work a hybrid filter wrapper model for identifying the disease affected children using the fMRI is proposed. An isolation tree filter model is used along with the newly proposed rider hawk optimization algorithm for wrapper feature extraction. Feature selection is a great challenge in machine learning where the unwanted variables are found and averted to improve the accuracy of the algorithm. By combining the power of filter and wrapper feature selection algorithms the proposed pipeline is tested on ABIDE dataset that has rs-fMRI images of the patient’s brain.  An accuracy of 77.67%  is obtained by using the proposed model for classification