Multi-Label Parkinson’s Disease Tremor Prediction using Ensemble Learning

Abstract

Parkinson\'s disease (PD) is a progressive neurodegenerative disorder for which early and accurate prediction is critical to improving patient outcomes. In this study, we propose an ensemble learning–based predictive framework that leverages the strengths of multiple classifiers to enhance diagnostic accuracy of PD detection. The model is trained and evaluated on a structured clinical dataset using optimized preprocessing and feature-selection techniques. To validate its effectiveness, the proposed ensemble model is compared with several widely used baseline classifiers, including Support Vector Machine (SVM), Decision Tree (DT), and K-Nearest Neighbors (KNN). Experimental results demonstrate that the ensemble learning approach achieves a prediction accuracy of 94%, outperforming SVM with accuracy(81.98%) , Decision Tree with accuracy(91.47%), and KNN with accuracy(76.23%) by substantial margins. This superior performance highlights the robustness, generalization capability, and enhanced discriminative power of ensemble methods for Parkinson’s disease prediction. The findings suggest that ensemble learning is a promising strategy for developing reliable, data-driven clinical decision-support systems for early PD detection.

Introduction

Parkinson’s disease (PD) is a progressive neurodegenerative disorder that is usually diagnosed late because traditional clinical assessments are subjective and depend on visible motor symptoms, by which time significant brain damage has already occurred. To enable earlier and more objective detection, recent research has increasingly focused on AI-based methods, including machine learning and deep learning models that analyze multimodal data such as speech, handwriting, MRI scans, wearable sensor signals, and clinical measurements.

Studies show strong performance of these approaches: wearable sensors with ML can track symptom progression, multimodal deep learning systems achieve very high accuracy by combining multiple data sources, and CNN-based MRI models and speech-based classifiers can detect PD with accuracies often above 90–98%. However, challenges remain in generalization, robustness across populations, and consistency of results.

To address these issues, the study proposes an ensemble learning–based predictive framework for early PD detection using structured clinical and tremor sensor data. The system uses the ALAMEDA tremor dataset, which contains 92 extracted features from wearable accelerometer signals and four binary labels representing different tremor types (rest, kinetic, postural, and constancy of rest). After preprocessing (imputation and normalization), the problem is treated as four separate binary classification tasks (one-vs-rest).

The model uses ensemble learning with boosted decision trees (MATLAB fitcensemble), where multiple weak learners are combined through boosting to improve accuracy and reduce errors. Feature selection and hyperparameter tuning are applied to enhance performance, and the dataset is split into 80% training and 20% testing.

Conclusion

The study demonstrated that the ensemble learning model developed using MATLAB’s fitcensemble provides superior performance for multi-label classification of Parkinson’s tremor types. It achieved the highest accuracy (94%) and the lowest Hamming Loss (0.059) compared to SVM, Decision Tree, and KNN, proving to be more robust, reliable, and suitable for clinical decision-support applications. These findings highlight the ensemble model’s robustness, reduced misclassification rate, and superior ability to generalize across diverse tremor patterns. The results emphasize the importance of combining weak learners to overcome limitations such as overfitting, instability, and sensitivity to noise that typically affect single classifiers. The enhanced discriminative performance makes the proposed ensemble framework highly suitable for real-world clinical decision-support systems, where early and reliable detection of PD tremor characteristics is critical. By accurately identifying subtle tremor signatures, the model has the potential to assist neurologists in early intervention, monitoring disease progression, and improving patient management.

References

[1] C. Sotirakis, Z. Su, M. A. Brzezicki, N. Conway, L. Tarassenko, J. J. FitzGerald, and C. A. Antoniades, “Identification of motor progression in Parkinson’s disease using wearable sensors and machine learning,” NPJ Parkinson’s Disease, vol. 9, no. 1, p. 142, Oct. 2023, doi: 10.1038/s41531-023-00581-2. [2] A. Sar, P. S. Puri, H. Naz, et al., “Multi-modal deep learning framework for early detection of Parkinson’s disease using neurological and physiological data for high-fidelity diagnosis,” Scientific Reports, vol. 15, p. 34835, 2025, doi: 10.1038/s41598-025-21407-6. [3] M. Alrawis, F. Mohammad, S. Al-Ahmadi, and J. Al-Muhtadi, “FCN-PD: An advanced deep learning framework for Parkinson’s disease diagnosis using MRI data,” Diagnostics, vol. 15, no. 8, p. 992, 2025, doi: 10.3390/diagnostics15080992. [4] T. H. H. Aldhyani, A. H. Al-Nefaie, and D. Koundal, “Modeling and diagnosis of Parkinson’s disease using hand drawing: Deep learning model,” AIMS Mathematics, vol. 9, no. 3, pp. 6850–6877, 2024, doi: 10.3934/math.2024334. [5] R. Alshammri et al., “Machine learning approaches to identify Parkinson’s disease using voice samples,” Frontiers in Artificial Intelligence, vol. 6, Mar. 2023, doi: 10.3389/frai.2023.1084001. [6] A. Mozhdehfarahbakhsh et al., “An MRI-based deep learning model to predict Parkinson’s disease,” medRxiv, Feb. 2021, doi: 10.1101/2021.02.19.21252081. [7] S. Rahman et al., “Classification of Parkinson’s disease using speech signal and machine learning/deep learning,” European Journal of Electrical and Computer Engineering, vol. 7, no. 2, 2023, doi: 10.24018/ejece.2023.7.2.488. [8] T. Tabashum et al., “Machine learning models for Parkinson disease: Systematic review,” JMIR Medical Informatics, vol. 12, 2024, doi: 10.2196/50117. [9] H. Rabie and M. A. Akhloufi, “A review of machine learning and deep learning for Parkinson’s disease detection,” Discover Artificial Intelligence, vol. 5, p. 24, 2025, doi: 10.1007/s44163-025-00241-9. [10] J. Fu, H. Chen, C. Xu, et al., “Harnessing routine MRI for the early screening of Parkinson’s disease: A multicenter machine learning study using T2-weighted FLAIR imaging,” Insights into Imaging, vol. 16, no. 1, p. 92, 2025, doi: 10.1186/s13244-025-01961-3. [11] B. Majhi, A. Kashyap, S. S. Mohanty, S. Dash, S. Mallik, A. Li, and Z. Zhao, “An improved method for diagnosis of Parkinson’s disease using deep learning models enhanced with metaheuristic algorithm,” BMC Medical Imaging, vol. 24, p. 156, 2024, doi: [12] 10.1186/s12880-024-01335-z. [13] N. U. Islam and R. Khanam, “Machine learning based Parkinson’s disease detection and progression evaluation using gray and white matter segmentation from 3D MRI,” Journal of Engineering and Applied Science, vol. 71, 2024. [14] S. Naimi et al., “1D-convolutional transformer for Parkinson disease diagnosis from gait,” arXiv preprint arXiv:2311.03177, 2023. [15] M. Hasan, “Parkinson’s Disease Freezing of Gait (FoG) Symptom Detection Using Machine Learning from Wearable Sensor Data,” arXiv preprint arXiv:2506.12561, 2025. [16] A. Nazir Mir et al., “Parkinson’s disease diagnosis through deep learning: A novel LSTM-based approach for freezing of gait detection,” arXiv preprint arXiv:2412.06709, 2024. [17] A. Li and C. Li, “Detecting Parkinson’s disease through gait measures using machine learning,” Diagnostics, vol. 12, no. 10, p. 2404, 2022, doi: 10.3390/diagnostics12102404. [18] G. P. Reddy et al., “Artificial intelligence-based effective detection of Parkinson’s disease using voice measurements,” Proceedings, vol. 82, no. 1, p. 28, 2024. [19] M. Al-Ameer et al., “Automated early prediction of Parkinson’s disease based on artificial intelligent techniques,” Arabian Journal for Science and Engineering, 2025. [20] P. Bhuvaneswari et al., “Detection of Parkinson’s disease using voice and spiral drawings on machine learning approaches,” Mahidol University Journal of Engineering and Technology, 2025. [21] I. Kalaiyarasi, P. Amudha, and S. Sivakumari, “Parkinson’s disease detection using deep learning technique,” International Journal for Research in Applied Science and Engineering Technology (IJRASET), vol. 11, no. 5, pp. 1250–1256, 2023. [22] M. Al-Khdour et al., “Detection of affected body parts in Parkinson’s disease and freezing of gait using time-series imaging and deep learning,” Sensors, vol. 23, no. 19, 2023.

Copyright

Copyright © 2026 Smita Ponde, Minakshi Rajput, Sharvari Tamane. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.