1University Department of Information Technology, University of Mumbai, Mumbai, India
Email: ppbhangale@yahoo.com
2University Department of Information Technology, University of Mumbai, Mumbai, India
Email: patilrajendrab@gmail.com
Autism Spectrum Disorder (ASD) is a complex disorder in neurodevelopment that should be screened early and correctly to facilitate a timely intervention and a better developmental outcome. The conventional diagnostic tests rely mainly on formal behavioural evaluations and specialist clinical opinion, which tend to be both time-intensive, subjective and unavailable. In order to address these constraints, this paper will present a multimodal expert system to predict ASD by combining structured behavioural screening data, unstructured texts reports by caregivers, and demographics as one machine learning system. This system uses naive natural language processing methods and compares various classification models, such as Naive Bayes, Support Vector Machines, Random Forest, XGBoost and deep learning models. Heterogeneous data modalities are combined with the help of feature fusion strategies, and class imbalance is solved by means of resampling and the loss-adjustment. To improve not only the ethical reliability but also the interpretability, fairness auditing and explainable artificial intelligence (XAI) techniques are added. The experimental data on publicly accessible Toddler and Adult ASD datasets show better performance compared to unimodal models, which proves the potential of the framework as a scalable and clinically interpretable decision-support system to detect ASD in the initial stages of its screening.
KEYWORDS: Autism Spectrum Disorder; Multimodal Expert System; Machine Learning; Naïve Natural Language Processing; Explainable AI; Fairness-Aware Prediction.
How to cite this article: Bhangale PP, Patil RB. Modelling, Implementation, and Performance Analysis of a Multimodal Expert System for Autism Spectrum Disorder Prediction. Int J Drug Deliv Technol. 2026;16(6s): 944-953; DOI: 10.25258/ijddt.16.6s.123
Source of support: None
Conflict of interest: None