Using machine learning to manage applied behavior analysis objectives in individualized education plans for children with autism spectrum disorder (2025)
- Authors:
- Autor USP: BOTELHO, BRUNO TEIXEIRA - ICMC
- Unidade: ICMC
- Subjects: APRENDIZADO COMPUTACIONAL; TRANSTORNO DO ESPECTRO AUTISTA; ENSINO INDIVIDUALIZADO; ANÁLISE DE SÉRIES TEMPORAIS
- Language: Inglês
- Abstract: This study addresses the clinical challenge of managing and adapting Individualized Education Plans (IEPs) for children with Autism Spectrum Disorder (ASD), a process reliant on time-intensive manual review. The objective was to investigate the feasibility of applying machine learning (ML) to predict the status progression ('Validated', 'Completed', or 'Rejected') of Applied Behavior Analysis (ABA) learning objectives, using longitudinal therapy data to create a foundation for an effective clinical decision-support tool. The methodology involved a longitudinal clinical dataset from a Brazilian healthcare provider, comprising 3,344 learning objectives from 438 children. A feature engineering process transformed the raw time series of progress scores into a set of 24 descriptive features, capturing the temporal dynamics of learning. A pre-trained Transformer-based model, Tabular Prior-Data Fitted Network (TabPFN), and a gradient-boosting model, XGBoost, were implemented and benchmarked against baseline algorithms, including Support Vector Machines (SVM), K-Nearest Neighbors (KNN), and a sequential Long Short-Term Memory (LSTM) network. The results demonstrated that TabPFN and XGBoost achieved superior performance with high overall accuracy (0.82 and 0.81, respectively), a weighted F1-score of 0.81, and weighted One-vs-Rest AUC-ROC scores of approximately 0.90. Both models effectively classified the 'Validated' (F1-score: 0.85) and 'Completed' (F1-score: ~0.80) classes, but struggled with the highly imbalanced 'Rejected' class. Feature importance analysis consistently identified the duration of the intervention and engineered features quantifying recent performance as the most influential predictors. The study concludes that ML models can effectively support the dynamic management of ABA IEPs by learning clinically relevant patterns from therapy data.This work provides a proof-of-concept for a clinical decision-support tool designed to augment, not replace, clinical expertise, paving the way for more scalable, responsive, and personalized interventions in ASD care and allowing therapists to focus on more complex clinical reasoning.
- Imprenta:
- Publisher place: São Carlos
- Date published: 2025
-
ABNT
BOTELHO, Bruno Teixeira. Using machine learning to manage applied behavior analysis objectives in individualized education plans for children with autism spectrum disorder. 2025. Trabalho de Conclusão de Curso (MBA) – Instituto de Ciências Matemáticas e de Computação, Universidade de São Paulo, São Carlos, 2025. Disponível em: https://bdta.abcd.usp.br/directbitstream/67bfbfef-1ae7-45f5-8091-ee79458834b6/Bruno_Teixeira_Botelho_TCC_2025.pdf. Acesso em: 27 fev. 2026. -
APA
Botelho, B. T. (2025). Using machine learning to manage applied behavior analysis objectives in individualized education plans for children with autism spectrum disorder (Trabalho de Conclusão de Curso (MBA). Instituto de Ciências Matemáticas e de Computação, Universidade de São Paulo, São Carlos. Recuperado de https://bdta.abcd.usp.br/directbitstream/67bfbfef-1ae7-45f5-8091-ee79458834b6/Bruno_Teixeira_Botelho_TCC_2025.pdf -
NLM
Botelho BT. Using machine learning to manage applied behavior analysis objectives in individualized education plans for children with autism spectrum disorder [Internet]. 2025 ;[citado 2026 fev. 27 ] Available from: https://bdta.abcd.usp.br/directbitstream/67bfbfef-1ae7-45f5-8091-ee79458834b6/Bruno_Teixeira_Botelho_TCC_2025.pdf -
Vancouver
Botelho BT. Using machine learning to manage applied behavior analysis objectives in individualized education plans for children with autism spectrum disorder [Internet]. 2025 ;[citado 2026 fev. 27 ] Available from: https://bdta.abcd.usp.br/directbitstream/67bfbfef-1ae7-45f5-8091-ee79458834b6/Bruno_Teixeira_Botelho_TCC_2025.pdf
Download do texto completo
| Tipo | Nome | Link | |
|---|---|---|---|
 |
Bruno_Teixeira_Botelho_TC... | Direct link |
How to cite
A citação é gerada automaticamente e pode não estar totalmente de acordo com as normas
