Detection of Autism Spectrum Disorder in Children Using Machine Learning Techniques

Area of Science:

• Neurology
• Developmental Pediatrics
• Computational Psychiatry

Background:

• Autism Spectrum Disorder (ASD) is a lifelong neurological condition impacting social, cognitive, and language skills.
• Early detection is crucial for improved outcomes, yet current diagnostic methods are time-consuming and costly.
• Machine learning offers potential to enhance ASD diagnostic precision and efficiency.

Purpose of the Study:

• To evaluate machine learning models for early Autism Spectrum Disorder (ASD) detection.
• To determine the most accurate predictive model for identifying ASD susceptibility in early developmental stages.
• To streamline the ASD diagnosis process through computational methods.

Main Methods:

• Applied machine learning algorithms including Support Vector Machines (SVM), Random Forest Classifier (RFC), Naïve Bayes (NB), Logistic Regression (LR), and KNN.
• Trained and evaluated predictive models on a dataset to assess ASD susceptibility.
• Compared the accuracy of different models for early ASD identification.

Main Results:

• Logistic Regression (LR) demonstrated the highest accuracy among the tested models for the selected dataset.
• The study successfully constructed predictive models for nascent stage ASD detection.
• Machine learning models show promise in complementing traditional ASD diagnostic tools.

Conclusions:

• Machine learning, particularly Logistic Regression, can significantly improve the accuracy and speed of early Autism Spectrum Disorder (ASD) detection.
• Implementing these models can aid clinicians in identifying children susceptible to ASD at earlier developmental stages.
• This approach has the potential to reduce diagnostic time and associated medical costs.