Multinomial logistic regression method for early detection of autism spectrum disorders

Abstract

A child with autism spectrum syndrome may have diverse personality traits, making it a heterogeneous condition, with cognitive abilities, emotional types, and so on. The system examined the child's behavioral data related to the diagnosis of autism spectrum disorder or developmental delay. The children aged from 6 months to 5 years who are identified as developmental delay has been clinically evaluated their behavioral assessment using Childhood Autism Rating Scale (CARS) tool. The main purpose of this article is to identify children with autism spectrum disorders using multinomial logistic regression. In this connection, this regression technique helps to identify the features of developmental delay or disorders, such as children's social relations, communication among the people and unrecognized activities. In this machine learning (ML) model training MLR-lbfgs produces 89%, MLR-sag provides 80%, MLR-saga provides 80%, and MLR-newton-cg provides best result of 97%. This tool is providing support to the clinicians for decision making process when unseeded autistic features or undecided situations. This tool is supported to unseen autistic features or undecided situations this tool provides best support to the clinicians to take decisions accurately and immediately.