Improving Detection of Autism Spectrum Disorder (ASD) by Using mRMR Feature Selection and Genetic Optimization Based CES Model for Computing Autism Severity Score

D. N. Disha, S. Seema, Archana Mathur

Abstract: Autism Spectrum Disorder (ASD) is one of the neurodevelopmental conditions that requires early and accurate diagnosis for effective intervention. The traditional screening tools such as Q-CHAT-10 and AQ-10 rely on predefined scoring mechanisms, which may not fully capture the variability in ASD symptoms. This study introduces a novel Genetic Optimization-Based CES Score (GOCES), an adaptive scoring framework that redefines ASD severity assessment. We used the mRMR algorithm (Minimum Redundancy Maximum Relevance) to identify the most significant behavioral characteristics from the Q-CHAT-10 and AQ-10 datasets. These selected features are subsequently used for building a GOCES model, where a Genetic Algorithm (GA) optimizes a new scoring function, the Constant Elasticity of Substitution (CES) production model. This approach dynamically adapts feature contributions, overcoming the limitations of fixed-score diagnostic tools. This research contributes a novel scoring paradigm that can turn the assessment into a new direction.

External IDs:dblp:journals/sncs/DishaSM25