Autism Spectrum Disorder Detecting Mechanism on Social Communication Skills Using Machine Learning Approaches

Abstract

Autism spectrum disorder alludes neuro-behavioral deficiency that involves sundry conventional autistic behaviors, and influences social circumstances of an autistic. It is intangible to heal completely, moreover it is complex due to the involvement of autistic demeanor, parents’ interlocution, and proficient medical adept. Alternatively, miscellaneous computational intelligence, and advanced methodologies of engineering such as computer vision, machine learning are dedicated to detect autism spectrum disorder in advance because it can be ameliorated by advance detection. However, superabundant number of implementations tolerate heterogeneity, imbalancing characteristics in data, severity in features, scarcity, and sparsity problem due to massive variations in datasets. This research has endeavored to eradicate all of those anomalies in autism spectrum disorder detection. The staple contribution is to detect the autism spectrum disorder in social communication skills of autistic characteristic participants using machine learning approaches. Specifically, autistic demeanor in social communication has been detected through this implementation. 10 basic features such as (AQ1–AQ10) have been carefully chosen, about 271 classifiers have been tested, and 9 among them have been determined as the best classifiers for detecting autistic characteristics in social communication skills. The propounded method has been applied on autistic characteristics based datasets repositories of toddlers, children, adolescent, and adult. The experimental results and evaluation section demonstrates that the proposed mechanism is comparatively effective, and performance results about 97.2% to 99.1% have been particularly determined with specific feature ranking evaluation.