Use of Machine Learning to Investigate the Quantitative Checklist for Autism in Toddlers (Q-CHAT) towards Early Autism Screening.

摘要: In the past two decades, several screening instruments were developed to detect toddlers who may be autistic both in clinical and unselected samples. Among others, Quantitative CHecklist for Autism Toddlers (Q-CHAT) is a quantitative normally distributed measure of traits that demonstrates good psychometric properties different settings cultures. Recently, machine learning (ML) has been applied behavioral science improve classification performance autism diagnostic tools, but mainly children, adolescents, adults. this study, we used ML investigate accuracy reliability Q-CHAT discriminating young children from those without. Five algorithms (random forest (RF), naive Bayes (NB), support vector (SVM), logistic regression (LR), K-nearest neighbors (KNN)) complete set items. Our results showed achieved an overall 90%, SVM was most effective, being able classify with 95% accuracy. Furthermore, using SVM–recursive feature elimination (RFE) approach, selected subset 14 items ensuring 91% accuracy, while 83% obtained 3 best common ours previously reported Q-CHAT-10. This evidence confirms high cross-cultural validity Q-CHAT, supports application create shorter faster versions instrument, maintaining as quick, easy, high-performance tool primary-care settings.