Applying Machine Learning to Autism Spectrum Disorder Diagnosis: Prediction and Classification Analyses
Tanmayi Panasa
09/02/2026
Autism Spectrum Disorder(ASD) is a prevalent neurodevelopmental condition that significantly impacts cognitive and social functions. Early and accurate diagnosis is crucial in providing appropriate care, intervention, and in mitigating future complications. However, ASD identification remains challenging due to its reliance on subjective behavioral assessments, symptom overlap with other disorders, and increased diagnostic difficulty with age. Poor awareness of ASD in the past is causing many adults to be diagnosed later in life. Undiagnosed adults are pressured to manage their symptoms without the necessary support. An accurate and objective diagnostic tool can bridge this gap. Considering these challenges, this study aims to enhance accurate diagnosis of ASD by identifying & training a reliable machine-learning model. Multiple supervised machine learning models, including Decision Tree, Random Forest, and Support Vector machine were trained and tested on an ASD screening dataset, containing 1,000 samples. Various programming elements, including Exploratory Data Analysis, Preprocessing, SMOTE, and hyperparameter tuning were taken into consideration developing these ML models. Their performance was evaluated based on cross-validation accuracy, accuracy score, and overall diagnostic reliability. The study revealed that Random Forest outperformed other models, with a cross-validation accuracy of 0.92, along with reliable accuracy, precision, and recall. By eliminating subjectivity, reducing bias, and improving affordability, machine learning presents a scalable and accessible approach to ASD identification. ML-based diagnosis can minimize manual analysis of screening exams, accelerating diagnostic processes and expanding access to a variety of populations.