Machine learning methods for autism spectrum disorder classification

It is important to acquire an early Autism Spectrum Disorder (ASD) diagnosis which can lead to early intervention. However, ASD assessment is based on behavioural measures which are usually subjective and cannot provide a diagnosis before the age of three. Autism research is characterised by interdi...

Full description

Saved in:
Bibliographic Details
Main Authors: Damianos, Lazaros, Vlachas, Christodoulos, Kollias, Konstantinos-Filippos, Asimopoulos, Nikolaos, Fragulis, George F.
Format: Conference Proceeding
Language:English
Subjects:
Artificial intelligence
Artificial neural networks
Autism
Classification
Diagnosis
Internet of Things
Machine learning
Neural networks
Robotics
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:It is important to acquire an early Autism Spectrum Disorder (ASD) diagnosis which can lead to early intervention. However, ASD assessment is based on behavioural measures which are usually subjective and cannot provide a diagnosis before the age of three. Autism research is characterised by interdisciplinarity, and Computer Engineering has contributed to it by offering technologies such as Robotics, Internet of Things (IoT) and Artificial Intelligence (AI). In the present study, Random Forests with Transfer Learning and Deep Neural Networks were implemented for autism detection employing the Autism Screening Adult Dataset acquired from the University of California, Irvine (UCI) Machine Learning Repository. Random Forests with Transfer Learning and Deep Neural Networks attained an ASD classification accuracy of 98.9% and 99.8%, respectively. Therefore, the method applied in this study showed that Machine Learning could offer promising methods for early, objective and accurate ASD classification compared to conventional autism assessment methods. Limitations and recommendations for future research are also included.
ISSN:0094-243X
1551-7616
DOI:10.1063/5.0182539