<p>Autism spectrum disorder (ASD) is one of the fastest growing neurodevelopmental disorders leading to life-long cognitive disabilities among children. Delayed diagnosis significantly hinders timely intervention, necessitating the development of intelligent early screening systems. Existing approaches to ASD detection predominantly rely on traditional screening methods that may lack precision and adaptability. The primary objective of the proposed system is to endow frontline health professionals with accurate and self-administered screening method with well-informed decision-making to aid fast referral decisions for ASD patients. The proposed study employs supervised machine learning (ML) hybrid classifiers with grid search-based hyperparameter tuning to construct a robust and efficient ASD screening framework. Four distinct ASD datasets, i.e., AQ-10 for child, adolescent and adult datasets and Q-CHAT-10 for toddlers are utilized, integrating patient data with a sentiment module that converts multiple choice questions into polarity scores reflecting social engagement and empathy traits. Sentiment-derived features are fused with numerical attributes to create a comprehensive feature space. Principal component analysis (PCA) is applied for feature extraction to enhance model generalization and predictive accuracy. The hybrid ML classifiers are rigorously evaluated using multiple performance metrics, demonstrating superior accuracy and reliability compared over existing ML-based approaches. The findings substantiate the effectiveness of the proposed framework in supporting early ASD detection and guiding clinical referral decisions.</p>