Enhancing Early Diagnosis of Autism in Toddlers: A Machine Learning Approach

The dramatic rise in Autism Spectrum Disorder (ASD) diagnoses, affecting roughly 1% of children globally according to the World Health Organization (WHO), necessitates a paradigm shift towards earlier detection. Current diagnostic methods often fail to identify ASD within the crucial window for early intervention, which can significantly improve a child's long-term prognosis. This study proposes a novel approach that leverages the power of Machine Learning (ML) to expedite ASD detection in toddlers. A comprehensive system is aimed to be developed that analyzes various data points, including behavioral patterns, language development, and physiological responses. By employing supervised learning algorithms, such as Logistic Regression and Random Forests, the system would be trained to identify subtle differences between autistic and typically developing children. Significant outcomes include: 1. The Support Vector Machine (SVM) algorithm demonstrated a 98% accuracy rate in ASD detection. 2. The Random Forest method exhibited superior performance, achieving 100% accuracy, precision, and recall. 3. Logistic Regression analysis yielded high accuracy levels, with 99.05% for training data and 98.10% for testing data. The initial explorations indicate that these algorithms hold significant promise in achieving this goal. This paves the way for the development of an accessible, online tool that empowers parents, caregivers, and healthcare professionals with the ability to conduct early and precise ASD screenings using machine learning. Imagine a future where this user-friendly platform could be accessed from anywhere in the world, potentially reducing diagnostic disparities and ensuring that every child with ASD receives the crucial support they need from the earliest possible stages. This could revolutionize the landscape of ASD intervention, leading to improved quality of life for countless individuals and their families.