A Review of Machine Learning Methods of Feature Selection and Classification for Autism Spectrum Disorder.

Area of Science:

• Neurodevelopmental Disorders
• Computational Psychiatry
• Artificial Intelligence in Healthcare

Background:

• Autism Spectrum Disorder (ASD) is characterized by social communication deficits and restricted behaviors, leading to social isolation.
• Early identification and intervention are vital for improving outcomes in children with ASD.
• Current diagnostic and therapeutic limitations, especially in rural areas, delay necessary support until school age.

Purpose of the Study:

• To review and analyze current machine learning (ML) methods for Autism Spectrum Disorder (ASD) feature selection and classification.
• To highlight the potential of ML in improving the speed, accuracy, and quality of ASD diagnosis.
• To recommend methods for enhancing ML's efficiency in processing complex ASD data.

Main Methods:

• Systematic review of recent studies on ML applications in ASD diagnosis.
• Analysis of various ML algorithms, including artificial neural networks, support vector machines, a priori algorithms, and decision trees.
• Focus on feature selection techniques crucial for predictive model development in ASD.

Main Results:

• Machine learning methods show potential for improving ASD diagnostic accuracy and reducing complexity.
• Feature selection is a critical step for developing effective ML-based predictive models for ASD.
• Various ML algorithms have been applied to ASD datasets to construct predictive models.

Conclusions:

• Advanced technologies like ML can significantly enhance early identification and intervention for ASD.
• Further research is needed to optimize ML for speedy execution and processing of complex, imbalanced ASD data.
• This review provides a foundation for future research utilizing ML in ASD diagnostics and data processing.