A COMPARATIVE STUDY ON AUTISM SPECTRUM DISORDER DETECTION BY LEVERAGING MACHINE LEARNING

A COMPARATIVE STUDY ON AUTISM SPECTRUM DISORDER DETECTION BY LEVERAGING MACHINE LEARNING

A.D.V.N. Murthy1, G. Neeraja2, G.Venkata Karthik3, B. Samanvitha4, G. Pavan Ganesh Naidu5

1 Assistant Professor

[2-5] B. Tech Student, LIET

[1,2,3,4,5] Computer Science & Engineering, Lendi Institute of Engineering and Technology, Vizianagaram

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract - We propose a study to compare machine learning (ML) algorithms for early detection of Autism Spectrum Disorder (ASD). Our approach involves four feature scaling methods and eight ML algorithms: Ada Boost, Random Forest, Decision Tree, K-Nearest Neighbours, XGBoost, Logistic Regression, Support Vector Machine, and Linear Discriminant Analysis. We evaluate these algorithms on datasets representing different age groups: Toddlers, Adolescents, Children, and Adults. After assessing performance using accuracy, precision, recall, F1-score, MCC, and Kappa, XGBoost emerges as the top performer, followed closely by Random Forest. Linear Discriminant Analysis and AdaBoost exhibit respectable scores, while Support Vector Machine and Logistic Regression offer moderate performance. K-Nearest Neighbors and Decision Tree perform weaker. Our study helps identify key factors contributing to ASD risk and ranks them using different techniques. This enables healthcare professionals to prioritize their assessments when screening for ASD. Overall, our method shows promise for early ASD detection, offering a valuable tool for healthcare providers.

Key Words:  Autism Spectrum Disorder, Machine Learning, Feature Scaling Technique, Feature Selection Technique, Classifications.