EARLY DETECTION OF AUTISM SPECTRUM DISORDER IN TODDLERS: A DATA-DRIVEN APPROACH

ABSTRACT Autism Spectrum Disorder (ASD) is a complex neurodevelopmental disorder characterized by difficulty in communication, social interaction, and behaviour that varies among individuals. It is critical to detect ASD early so it can be treated on time but toddlers with ASD are hard to diagnose as the symptoms can differ greatly and overlap with other developmental disorders. This research proposes using machine learning to diagnose ASD at an early stage . This is done through a multimodal dataset that consists of standard screening tool information. Specifically, they will cover Q-CHAT-10, ADOS, and INCLEN in detail. The dataset which is used to either classify someone or not someone to have autism is from the social interaction, communication, sensory and repetitive behaviour domain. By incorporating various behavioral markers, the data’s multimodal aspect improves the model’s capability of detecting subtle patterns that indicate ASD. To find out which classifier gives the best results, different machine learning algorithms were evaluated on this dataset. The different algorithms which were evaluated include Logistic Regression, Support Vector Machine, Naive Bayes, Decision Tree, Random Forest, AdaBoost, Gradient Boosting, Bagging, Deep Neural Networks, etc. To measure the performance of the various models, accuracy, precision, recall, F1-score, etc. were used. The findings showed that multiple models especially Naive Bayes, Logistic Regression, Random Forest and Deep Neural Networks have high values of accuracy and recall and therefore have a reliable capacity to distinguish between the ASD-positive and ASD-negative cases. This study shows how machine learning can accurately diagnose toddlers with ASD through a multimodal model. Combining various diagnostic tools and further evaluating a range of classification algorithms contribute to creating a robust data-driven approach. Therefore, this study offers an intervention to clinicians and parents for early-stage identification of ASD (Autism Spectrum Disorder). This paper analyzes the detection of Autism Spectrum Disorder using classifiers on multimodal data. Keywords: Autism Spectrum Disorder, early detection, machine learning, multimodal dataset, classifier comparison, Q-CHAT-10, ADOS, INCLEN.