The Genetic-Evolutionary Random Support Vector Machine Cluster Analysis In Autism Spectrum Disorder

Xia-an Bi,Yingchao Liu,Qi Sun,Xianhao Luo,Haiyan Tan,Jie Chen,Nianyin Zeng

DOI: https://doi.org/10.1109/ACCESS.2019.2902889

IF: 3.9

2019-01-01

IEEE Access

Abstract:Previous researches have produced a number of conclusions on the functional magnetic resonance imaging (fMRI) study for autism spectrum disorder (ASD) patients, but there are different opinions about the brain regions of the lesions. In order to study ASD more deeply, an advanced framework, i.e., genetic-evolutionary random support vector machine (SVM) cluster, was proposed in this paper. In our method, an initial cluster of multiple SVMs was first built by randomly picking samples and features. Then, these SVMs were selected to recombine and mutate the aim of genetic evolution until the number of genetic evolution which reached the threshold or the classification accuracy was stable. We evaluated the proposed method by using the resting state fMRI data (103 ASD patients and 106 healthy controls), which achieved a 96.8% accuracy. Based on the classification results, the abnormal brain regions were found out. This study suggests the pathogenesis of ASD to a certain extent and offers great assistance for the diagnosis of potential patients with ASD.