Parkinsons Disease Speech Dataset

Dataset Overview

Dataset Name

A Machine Learning Approach for the Diagnosis of Parkinson's Disease via Speech Analysis

Research Time

March 2022

Dataset Source

University of Oxford

Dataset Composition

• Number of Instances: 195
  • 147 Parkinson's subjects
  • 48 without Parkinson's
• Number of Features: 22
  • Includes features such as frequency, pitch, amplitude/period of the waveform, etc.
• Label: 1 represents Parkinson’s, 0 represents non‑Parkinson’s

Algorithms Used

1. Logistic Regression (LR)
2. Linear Discriminant Analysis (LDA)
3. k Nearest Neighbors (KNN)
4. Decision Tree (DT)
5. Neural Network (NN)
6. Naive Bayes (NB)
7. Gradient Boost (GB)

Engineering Goal

Develop a machine‑learning model for Parkinson’s diagnosis, achieving at least 90 % accuracy and/or a Matthews Correlation Coefficient of at least 0.9.

Data Analysis Results

After rebalancing the dataset, a 75‑25 train‑test split yielded the best performance. K‑Nearest Neighbors and Neural Network achieved a maximum accuracy of 98 %.

Conclusion

The project demonstrates that machine learning significantly improves Parkinson’s diagnosis compared with existing methods, achieving 98 % accuracy, which is crucial for effective treatment.