Predicting Autism Spectrum Disorder in Children Using Glowworm Optimization With Extreme Learning Machine Networks

Area of Science:

• Pediatric Healthcare
• Machine Learning Applications
• Developmental Disorders

Background:

• Early prediction of autism spectrum disorder (ASD) is crucial for timely intervention, improving developmental outcomes.
• Current ASD detection methods are often time-consuming, subjective, and lack accessibility, particularly in rural areas.
• Developing scalable, objective, and rapid ASD detection systems is essential to overcome healthcare challenges.

Purpose of the Study:

• To enhance the efficiency and accuracy of autism spectrum disorder (ASD) prediction.
• To address limitations in current ASD detection processes, including time constraints and accessibility issues.
• To provide a reliable and scalable solution for early ASD identification in pediatric healthcare.

Main Methods:

• Integration of the Glowworm Optimization with Extreme Learning Machine Networks (GO-ELMN) model for ASD prediction.
• Utilizing behavioral, demographic, and medical features from children's ASD screening data.
• Optimizing network hyperparameters with the glowworm optimization algorithm to handle limited and imbalanced data.

Main Results:

• The GO-ELMN model demonstrated high accuracy in ASD prediction.
• The system achieved a fast convergence speed, indicating computational efficiency.
• Experimental results validated the system's effectiveness in identifying children's behavior patterns.

Conclusions:

• The developed ASD detection model offers an interpretable, fast, and reliable solution.
• The system is suitable for effective utilization within the pediatric healthcare domain.
• This approach addresses key challenges in early autism spectrum disorder identification.