Advancing Autism Diagnosis: The Role of Machine Learning in Overcoming Traditional Barriers

Overview

Autism Spectrum Disorder (ASD) is a complex neurodevelopmental condition whose diagnosis remains highly dependent on clinician expertise, time-intensive behavioral assessments, and limited accessibility—particularly for adults and underrepresented populations. This study investigates the potential of machine learning (ML) to enhance ASD screening using structured questionnaire data from the UCI Autism Screening Adult Dataset (n = 704, 17 features). Four supervised ML algorithms—Logistic Regression, Support Vector Machine (SVM), Neural Network, and Random Forest—were trained and optimized through five-fold cross-validation. Feature correlation analysis and interpretability tools, including SHAP and permutation importance, were employed to identify the most discriminative predictors. Random Forest and Neural Network models achieved the highest performance (accuracy > 99%), while SVM and Logistic Regression provided slightly lower but more generalizable results. Across all models, three behavioral features—A9 (“facial expression recognition”), A5 (“reading between the lines”), and A6 (“interpreting boredom cues”)—consistently emerged as the strongest predictors of ASD, whereas demographic variables contributed minimally. Dimensionality-reduction analyses (PCA and t-SNE) further confirmed clear separability between ASD-positive and negative groups. These findings demonstrate that accurate, transparent ASD screening can be achieved using a concise subset of questionnaire items, supporting the development of lightweight, accessible diagnostic tools for early identification and intervention.

Advancing Autism Diagnosis: The Role of Machine Learning in Overcoming Traditional Barriers

Overview

Competition history

Related projects

Optimizing Random Forest Algorithms for Autism Prediction: Enhancing Diagnostic Accuracy Through Machine Learning

Towards Improved Recognition and Diagnosis of Autism Among Females – A Novel Approach Using Machine Learning

Autism Spectrum Disorder Screening Using Machine Learning

Identifying Key Factors to Improve Autism Spectrum Disorder Diagnosis with Machine Learning

Deep Learning for Autism Diagnosis: Insights into Genomic and Phenotypic Features

Towards Improved Recognition and Diagnosis of Autism in Females

Diagnosing Autism with Machine Learning: Binary Classification for Eye Movement in Virtual Reality Environment

Identification of Neurodevelopmental & Mental Health Conditions: An AI Approach to Reducing Barriers

Browse more like this