Zaibunnisa L.H. Malik* and Pooja Raundale
The proportion of children with ASD at risk for a motor impairment was very high at 86.9%. Children with ASD did not outgrow their motor impairments and continued to present with a risk for Developmental Coordination Disorder (DCD) even into adolescence. Yet, only 31.6% of children were receiving physical therapy services. To diagnose ASD, the clinical standardized tests are being used. To test and predict ASD the lengthy diagnostic time is required and also these tests are very costly. Machine learning techniques are being used in place of traditional methods to reduce the time and cost required to predict ASD diagnose. The field of Machine Learning, a branch of Artificial Intelligence, is utilized to diagnose Autism Spectrum Disorder (ASD) by treating it as a classification task. The proposed techniques are evaluated on six different non-clinically ASD datasets. We have applied models such as Logistic Regression (LR), Support Vector Machine (SVM), KNearest Neighbors (KNN), Random Forest Classifier (RFC), Neural Network (NN) to our dataset and created predictive models based on the outcome. The main objective of our paper is to determine if the teenager is likely to have ASD or not. The prediction model will help the doctors to streamline the diagnosis process. Based on our results, Logistic Regression (LR) achieved better as compared to other algorithms, with an accuracy close to 95.84%. This research paper focuses on the application of machine learning techniques and the assessment of their effectiveness on an autism spectrum disorder dataset. The results include a comparative analysis of various algorithms for ASD diagnosis.
Published Date: 2025-03-19; Received Date: 2023-11-05
