Applying machine learning to eye-tracking data for autism identification in high-functioning adults

Individuals with high-functioning Autism Spectrum Disorder (ASD), lacking intellectual disabilities and exhibiting subtle functional, not easily conspicuous deficits, are frequently diagnosed with autism during adulthood. Early and objective autism screening and intervention can enhance many aspects of their and their families’ lives as it is the first vital step in tackling with this situation. Current ASD research fosters the implementation of cutting-edge assessment tools, i.e., Machine Learning (ML), eye-tracking technology, robotics, Internet of Things (IoT) etc., rather than relying primarily on subjective behavioural assessment instruments. The present study employed a dataset from a prior study, comprising eye-tracking data from high-functioning autistic adults engaged in a Browse and a Search web-related tasks. MATLAB and various Machine Learning classification algorithms, such as Neural Networks, Decision Trees, Support Vector Machines, Logistic Regression, Naive Bayes, and an Ensemble model were utilised. The highest ASD classification test process results were 82.4% in the Browse task and 85.1% in the Search one, when Neural Networks were implemented. The high classification accuracy achieved underscores the potential of applying Machine Learning on eye-tracking data for early and objective autism detection.