Detecting Parkinson's Disease Using Vocal Features

The report explores using machine learning to predict Parkinson’s Disease based on vocal features, aiming for non-invasive early diagnosis.It details preprocessing steps like handling class imbalance with SMOTE and standardizing features, followed by EDA revealing key vocal differences in patients.Multiple models (Random Forest, SVC, Gradient Boosting) were evaluated, with Random Forest achieving the highest accuracy (94%) and performance metrics.Findings highlight vocal frequency and amplitude variations as critical predictors, supported by visualizations like boxplots and correlation heatmaps. https://github.com/alihassan098799/datastewardship_1本报告探讨了基于语音特征的机器学习帕金森病（Parkinson’s Disease）预测方案，旨在实现无创早期诊断。报告详述了数据预处理流程：通过SMOTE（合成少数类过采样技术）处理类别不平衡问题，并对特征进行标准化；随后借助探索性数据分析（Exploratory Data Analysis，EDA）揭示了患者群体的关键语音特征差异。研究评估了多种机器学习模型，包括随机森林（Random Forest）、支持向量分类机（Support Vector Classification, SVC）与梯度提升树（Gradient Boosting），其中随机森林取得了最高的准确率（94%）与最优的综合性能指标。研究结果凸显了语音频率与振幅的变化作为核心预测因子的重要性，该结论通过箱线图、相关热图等可视化工具得以佐证。https://github.com/alihassan098799/datastewardship_1