스태킹 앙상블 기반 머신러닝 모델을 활용한 파킨슨병 보행데이터 해석 연구
A Study on the Interpretation of Parkinson’s Disease Gait Data Using Stacking Ensemble-Based Machine Learning Models
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
This study proposes an interpretable machine learning model to assist in the early-stage classification of Parkinson’s disease, aiming to support early diagnosis. A stacking ensemble approach integrating Support Vector Machine (SVM), Random Forest, and XGBoost classifiers was employed on gait data obtained from the PhysioNet Gait in Parkinson’s Disease dataset, consisting of 93 patients (mean age 66.3±10.9 years, Hoehn & Yahr stage 2–3) and 73 healthy controls. Feature standardization and recursive feature elimination with cross-validation (RFECV) were applied to select informative gait features. The proposed model achieved 96.6% classification accuracy and an AUC of 0.990, demonstrating strong generalization and interpretability. These findings suggest that the proposed stacking ensemble framework could serve as a foundation for future predictive models for early Parkinson’s detection.
Abstract
This study proposes an interpretable machine learning model to assist in the early-stage classification of Parkinson’s disease, aiming to support early diagnosis. A stacking ensemble approach integrating Support Vector Machine (SVM), Random Forest, and XGBoost classifiers was employed on gait data obtained from the PhysioNet Gait in Parkinson’s Disease dataset, consisting of 93 patients (mean age 66.3±10.9 years, Hoehn & Yahr stage 2–3) and 73 healthy controls. Feature standardization and recursive feature elimination with cross-validation (RFECV) were applied to select informative gait features. The proposed model achieved 96.6% classification accuracy and an AUC of 0.990, demonstrating strong generalization and interpretability. These findings suggest that the proposed stacking ensemble framework could serve as a foundation for future predictive models for early Parkinson’s detection.