Machine Learning Methods for the Timely Identification of Autism Spectrum Disorder in Toddlers

Abstract

Autism Spectrum Disorder (ASD) is an evolving ailment that upsets the nervous system, by means of a widespread range of symptoms and severity levels. Early ASD identification is challenging and there is no treatment, early detection and intervention are vital for improving outcomes in individuals with ASD. This research, explored the application of machine learning (ML) techniques, specifically Support Vector Machines (SVM), Random Forest (RF), and Gradient Boosting (GB), for detection of ASD using a dataset focused on screening toddlers. The dataset, obtained from Kaggle in the 2018 release, encompasses demographic information, medical history, and screening test results for toddlers. The presented system employs leading edge algorithms to analyze the dataset and build predictive models for ASD detection. SVM is preferred due to its capability to lever multifaceted associations in high-dimensional data, while RF excels in handling diverse feature sets and capturing nonlinarites.GB, known for its ensemble learning capabilities, is also employed to boost predictive performance. The assessment of these models is accomplished via bench mark estimation metrics in ML. This approach aims at an effective model for ASD detection in children based on the characteristics of the dataset. Additionally, feature importance analysis is conducted to gain insights into the factors contributing to ASD prediction. The implementation results reveals that SVM algorithm outperforms in the proposed ASD diagnosis. Early identification using ML techniques could assist healthcare professionals and caretakers in timely intervention, ultimately improving the prolonged effects for persons with ASD.