Diagnosing autism spectrum disorders using a double deep Q-Network framework based on social media footprints

Area of Science:

• Neuroscience
• Computational Linguistics
• Artificial Intelligence

Background:

• Social media is a growing tool in healthcare for rapid information dissemination.
• Autism Spectrum Disorder (ASD) presents unique challenges and opportunities for digital engagement.
• Twitter serves as a significant platform for the ASD community to share information and experiences.

Purpose of the Study:

• To analyze Twitter data for identifying Autism Spectrum Disorder (ASD) using machine learning.
• To examine behavioral traits and emotional expressions of individuals with ASD on social media.
• To evaluate the efficacy of AI models in detecting ASD through online interactions.

Main Methods:

• Utilized Twitter as the primary data source for analyzing ASD-related content.
• Applied advanced AI models including CNN-LSTM, LSTM, and a DDQN-Inspired approach.
• Preprocessed tweet data by cleaning text, tokenizing, and encoding for binary classification.

Main Results:

• The DDQN-Inspired model achieved a high precision rate of 87% in identifying ASD.
• The proposed approach demonstrated significant potential for ASD identification from social media content.
• Comparison with existing systems highlighted the effectiveness of the developed framework.

Conclusions:

• The developed system can assist clinicians in studying ASD symptoms within digital footprints.
• Analyzing social media text variations offers a novel method for ASD-related research.
• AI-driven analysis of social media holds promise for understanding and supporting individuals with ASD.