Enhancing ASD detection accuracy: a combined approach of machine learning and deep learning models with natural...

Area of Science:

• Computational linguistics
• Medical informatics
• Neuroscience

Background:

• Autism Spectrum Disorder (ASD) diagnosis presents challenges, including the need for specialized professionals and extensive resources.
• Early identification of ASD is crucial for timely intervention and improved quality of life, especially in children.

Purpose of the Study:

• To explore the utility of artificial intelligence (AI), specifically machine learning (ML) and deep learning (DL) models, in diagnosing Autism Spectrum Disorder (ASD).
• To analyze text data from social media platforms like Twitter for potential ASD case detection, overcoming diagnostic resource limitations.

Main Methods:

• Employed natural language processing (NLP) techniques and various ML/DL models (Decision Trees, XGB, KNN, RNN, LSTM, Bi-LSTM, BERT, BERTweet).
• Utilized a dataset of 404,627 tweets from Twitter, with a subset of 90,000 tweets (45,000 ASD, 45,000 non-ASD) for model training and testing.

Main Results:

• AI models demonstrated promising performance in classifying texts potentially from individuals with ASD.
• The predictive model achieved an accuracy of nearly 88% in identifying potential ASD cases from text data.

Conclusions:

• AI, particularly DL models, can significantly enhance the accuracy of ASD detection and diagnosis.
• This approach highlights AI's potential in advancing early diagnostic techniques for better patient outcomes and emphasizes the importance of early ASD identification.