Scholarlink

<p>The research discusses an innovative hybrid machine learning model used for detecting Parkinson’s disease based on voice measurement inputs. A diagnostic improvement method will integrate features and dimension reduction while implementing ensemble classification approaches. The implementation of SelectKBest leads to feature selection which is followed by Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) procedures for data reduction. The framework incorporates four diverse classifiers: It was also optimized through GridSearchCV hyperparameter tuning of K-Nearest Neighbors (KNN), MLP, Random Forest and XGBoost. Our framework is evaluated on a dataset of 195 voice recordings of 31 individuals. Our hybrid approach is found to outperform baselines, evidenced by the high performance of the PCA augmented KNN model with 97.44% accuracy at classifying the ones with and without Parkinson’s disease. This is further supported by ablation studies which show that both feature selection as well as dimensionality reduction have pinned down improvement of the system’s performance. An analysis of the confusion matrix shows that the model is correct to identify Parkinson’s cases almost without error. Overall, this research presents the method of hybrid machine learning frameworks for improving early Parkinson’s disease detection accuracy and robustness and contribute to field of medical diagnostics. This method demonstrates potential application across different biomedical classification tasks to expand the diagnosis capabilities within medical domains.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Hybrid Machine Learning Framework for Parkinson’s Disease Detection: Integrating Feature Selection, Dimensionality Reduction, and Ensemble Classification

  • 期刊论文

  • 发表日期：

    2026年2月7日

查看全文
  • T. S. Chandrakantha,
  • Basavaraj N. Jagadale,
  • G. R. Madhuri

摘要

The research discusses an innovative hybrid machine learning model used for detecting Parkinson’s disease based on voice measurement inputs. A diagnostic improvement method will integrate features and dimension reduction while implementing ensemble classification approaches. The implementation of SelectKBest leads to feature selection which is followed by Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) procedures for data reduction. The framework incorporates four diverse classifiers: It was also optimized through GridSearchCV hyperparameter tuning of K-Nearest Neighbors (KNN), MLP, Random Forest and XGBoost. Our framework is evaluated on a dataset of 195 voice recordings of 31 individuals. Our hybrid approach is found to outperform baselines, evidenced by the high performance of the PCA augmented KNN model with 97.44% accuracy at classifying the ones with and without Parkinson’s disease. This is further supported by ablation studies which show that both feature selection as well as dimensionality reduction have pinned down improvement of the system’s performance. An analysis of the confusion matrix shows that the model is correct to identify Parkinson’s cases almost without error. Overall, this research presents the method of hybrid machine learning frameworks for improving early Parkinson’s disease detection accuracy and robustness and contribute to field of medical diagnostics. This method demonstrates potential application across different biomedical classification tasks to expand the diagnosis capabilities within medical domains.