Hybrid Machine Learning Based Model for Early Parkinson's Disease Detection

Sotelo R.; Perez C.; Santisteban J. · International Conference on Artificial Intelligence, Computer, Data Sciences, and Applications, ACDSA 2026 (2026) · DOI: 10.1109/acdsa67686.2026.11467613

Resumen

Early detection of Parkinson's disease remains challenging due to the absence of accessible biomarkers and objective diagnostic tools. This study presents a hybrid machine learning model that combines structured data and graphical biomarkers through a multimodal late-fusion approach. The proposed method combines Gradient Boosting and Convolutional Neural Network (CNN) models, achieving improved interpretability and computational efficiency compared to attention-based fusion models. Results show that Gradient Boosting achieved 95.25% accuracy, while CNN achieved 97.40% accuracy. Their adaptive late fusion attained 98.27% accuracy and a 96.6% F1-score. A 5-fold crossvalidation confirmed robustness and absence of overfitting, demonstrating potential for cost-effective and reproducible early diagnosis of Parkinson's disease in clinical contexts across Latin America.

Instituciones: Universidad Peruana de Ciencias Aplicadas