Table 1.

Comparison between previous studies on autism spectrum disorder (ASD) diagnosis and intervention and the present study.

Authors	Year	Contributions	Differences from Present Study
Lord and Luyster [5]	2006	Demonstrated stability of ASD diagnosis from age 2 to 9; emphasized reliability of early diagnosis.	Does not incorporate ML or XAI; focuses on longitudinal stability rather than predictive modeling.
McCarty and Frye [6]	2020	Identified challenges in early diagnosis; proposed multi-stage screening to improve diagnostic accuracy.	Does not utilize ML algorithms or XAI techniques; focuses on screening methodologies.
Bryson et al. [7]	2003	Reviewed impact of early intervention on developmental outcomes; emphasized importance of early detection tools.	Lacks application of ML models and XAI; focuses on intervention strategies rather than predictive analytics.
Guthrie et al. [8]	2013	Examined stability of early diagnoses; highlighted need for multifaceted diagnostic approaches.	Does not apply ML or XAI; emphasizes clinical expertise and multi-source information integration.
Omar et al. [12]	2019	Developed ML models using RF-CART and RF-ID3; created a mobile diagnostic application for ASD.	Focuses on model development and application without integrating XAI techniques; limited interpretability of model predictions.
Usta et al. [13]	2019	Evaluated ML algorithms for predicting short-term ASD outcomes; found decision tree to be most effective.	Does not incorporate XAI for model interpretability; primarily focuses on prognosis prediction rather than diagnostic accuracy.
Alsuliman and Al-Baity [26]	2022	Developed optimized ML models for ASD classification using PBC and GE data with bio-inspired algorithms.	Does not integrate advanced XAI techniques to improve both the accuracy and interpretability of their models.
Ben-Sasson et al. [27]	2023	Developed a gradient boosting model for early ASD prediction using electronic health records.	Lacks integration of XAI techniques; focuses on predictive modeling using electronic health records.
Abbas et al. [28]	2023	Compared TPOT and KNIME for ASD detection, focusing on feature selection.	Does not include XAI techniques; focuses on comparing AutoML tools for ASD detection.
Reghunathan et al. [29]	2023	Used machine-learning classifiers for ASD detection, with logistic regression showing the highest accuracy.	Does not use XAI techniques; focuses on feature reduction and classifier accuracy.
Bala et al. [30]	2023	Built an ASD detection model across age groups, with SVM performing best.	Focuses on model performance across age groups without applying XAI for interpretability.
Batsakis et al. [31]	2023	Built a data-driven AI model for clinical ASD diagnosis, highlighting data limitations.	Emphasizes model development using AutoML without integrating XAI for improved interpretability.
Our Study	2024	Developed interpretable ML models using XAI techniques; implemented rigorous data preprocessing; provided guidelines for non-experts.	Integrates XAI for model transparency; emphasizes data reliability and preprocessing; offers practical guidelines for clinical use.