Data-centric automated approach to predict autism spectrum disorder based on selective features and explainable...

Area of Science:

• Neuroscience
• Computer Science
• Education

Background:

• Autism spectrum disorder (ASD) is a neurodevelopmental condition impacting cognitive function, language, social interaction, and communication.
• Early identification and intervention for ASD can significantly reduce the need for extensive medical treatments and lengthy diagnostic processes.
• Genetic factors are the primary cause of ASD, highlighting the importance of precise diagnostic tools.

Purpose of the Study:

• To develop a machine learning model for early prediction of autism spectrum disorder (ASD) in toddlers.
• To identify and propose tailored educational strategies for children with ASD based on their unique behavioral, verbal, and physical responses.
• To enhance the accuracy and efficiency of ASD diagnosis and intervention through advanced computational methods.

Main Methods:

• Utilized three feature engineering techniques: Chi-square, backward feature elimination, and Principal Component Analysis (PCA).
• Applied multiple machine learning models, including the proposed XGBoost 2.0, for ASD presence prediction in toddlers.
• Assessed behavioral, verbal, and physical responses to identify personalized educational methods for children with ASD.
• Implemented cross-validation to ensure model stability and compared results with previous research.

Main Results:

• The proposed XGBoost 2.0 model achieved 99% accuracy, F1 score, and recall, with 98% precision using chi-square significant features for ASD prediction.
• The approach demonstrated 99% accuracy, F1 score, recall, and precision in identifying tailored educational methods for children with ASD.
• Cross-validation confirmed the stability of the proposed models.

Conclusions:

• Machine learning, particularly XGBoost 2.0, shows high efficacy in predicting ASD in toddlers.
• The study successfully developed a framework for creating personalized educational strategies for individuals with ASD.
• This research offers a significant advancement in utilizing computational techniques for improved ASD diagnosis and tailored educational interventions.