Abstract
Autism Spectrum Disorder (ASD) is a mental disorder characterized by difficulties with socializing, repetitive behaviors, speech and non-verbal communication. It is diagnosed within the first three years of life. The earlier the diagnoses, the sooner the intervention can start. Previous studies showed that early interventions for ASD children result in higher success rate. Thus, early diagnosis is an important research goal. Clinical instruments currently used for measuring ASD signs and symptoms are time consuming and highly influenced by subjective observations. These limitations delayed diagnosis and further intervention. Therefore, scientists found that atypical gaze movement is among the earliest biomarkers for ASD. Accordingly, in our research, we are aiming to speed up diagnoses by combining gaze-based screening with intelligent methods such as machine learning which would act as a transformative step for identifying ASD at early stages. In this research we used Support Vector Machine (SVM) algorithm to examine the performance in terms of four different measures which are accuracy, sensitivity, specificity and Area under the curve (AUC). Results revealed that SVM accomplished high classification performance when applied on our collected eye movement data set.