Children with autism spectrum disorder (ASD) have difficulties communicating and interacting with others, exhibit restricted and repetitive activities, and have trouble learning new things. Since behavioural screening is the initial stage in clinically diagnosing ASD, early intervention is delayed. The electroencephalography (EEG) technique has been helpful in the identification of several neurological disorders by examining the electrical activity of the brain. A feature called Slope Entropy (SlopEn) is employed to measure the properties derived from EEG data. Patients are categorised into mild, moderate, and severe groups using support vector machines (SVM) and random forests (RF). The main objective of this study is to use SlopEn to identify a biomarker for ASD that can be used to assess the severity of moderate, mild, and severe patients using a two-way ANOVA among the brain regions. A secondary goal is to suggest an automatic model that uses SVM and RF machine learning classifiers to detect the ASD severity from typically developing children with SVM accuracy of 98.67%, while the RF-based classification has 97.89% accuracy.