Ensemble Machine Learning for Classification of Autism Spectrum Disorder in Toddlers and Adults

Autism Spectrum Disorder (ASD) is a neurological disorder. A Person with ASD always faces challenges in communicating socially and is involved in repetitive behaviors. Accurate and timely diagnosis is essential for efficient assistance and action. This paper presents the ensemble machine learning methodology to classify ASD in both toddlers and adults. Four distinct algorithms, Gradient Boosting (GB), Histogram Boosting (HB), Extreme Gradient Boosting (XGB) and Adaptive Boosting (ADB), are used. Explorative data analysis is performed to show the impact of behavioural features and individual characteristics on ASD in toddlers and adults. Quantitative analysis demonstrates that XGB outperforms with classification accuracy, log loss and F1-score, precision and recall. The findings indicate that the ensemble machine learning methodology has great potential to improve the diagnostic procedures for ASD, possibly resulting in an earlier and more accurate diagnosis of the condition.