Intelligent Autism Spectrum Disorder Diagnosis: A Hybrid Ensemble Learning Framework

Autism spectrum disorder (ASD) affects the agerelated skills in children, such as cognitive and social skills. Due to this, children face complications in social interaction, limited interests, and repetitive actions. Early detection of ASD can prevent major effects and long-term disorders. This paper proposed an ensemble learning classifier to identify the autism disorder presence in the ASD dataset. The dataset has both categorical and numerical variables, which is collected from the different groups of children in the range of $\mathbf{1 2}$ to $\mathbf{3 6}$ months. The ensemble learning method is initially trained with different machine learning (ML) classifiers and combines the bestperforming ML models as an ensemble learning technique to improve performance. Support vector machine (SVM), decision tree (DT), and random forest (RF) classifiers were the best performers during experimentation, and later they were combined as an ensemble learning method using soft voting. The performance of the proposed ensemble learning classifier is compared with different machine learning models and the proposed approach recorded the highest accuracy compared to others. A receiver operating characteristic (ROC) curve is provided to analyze and validate the performance of the proposed ensemble learning method.