Early Prediction of Autism Spectrum Disorder by Deep Learning and Ensemble Learning

Abstract

Neurodevelopmental condition of the person which affects how a human interact, communicate and perceives the persons in the globe. This process is called as spectrum. People communicating with other persons repeated words in behavior is called Autism spectrum Disorder(ASD). Generally it begins in children and need early monitoring otherwise it may cause serious loss to human. Limited Eye contact, No importance for care takers and early infancy are the symptoms of ASD. Proposed research used AlexNet and RegNet from the Deep Learning(DL), Gradient Boosting and Extra Trees Classifier(ETC) from the ensemble learning for ASD identification. ETC from the ensemble learning produces highest accuracy(92%) to identify ASD over gradient boosting. AlexNet from the DL gets highest accuracy(98%) to identify ASD over RegNet. AlexNet DL model is to obtain abnormalities and shows highly accurate in ASD classification over conventional techniques. Early Monitoring can save patient health including disease spreading to other parts of the organ.