Parksinson's Disease Detection
Not the Neurons in ML models
Parkinson’s Disease Detection is a machine learning project focused on identifying Parkinson’s disease using voice-based biomedical features. The objective is to train and evaluate multiple supervised and unsupervised learning models and compare their performance in predicting the disease.
The project uses a dataset containing 195 samples and 24 features, where the status column serves as the dependent variable indicating the presence or absence of Parkinson’s disease. The features include frequency, jitter, shimmer, noise ratios, and non-linear measures extracted from voice signals.
Several models were implemented and analyzed. Supervised learning models include Random Forest, Logistic Regression, Decision Tree, K-Nearest Neighbour, Support Vector Machine, Perceptron, Gaussian Naive Bayes, LightGBM, and XGBoost. Unsupervised learning was explored using K-Means Clustering to understand underlying data patterns.
The best-performing model, Support Vector Machine achieved a maximum accuracy of over 80%, demonstrating the effectiveness of machine learning techniques for Parkinson’s disease detection.
Overall, the project emphasizes model comparison, problem-solving, and practical application of machine learning techniques for healthcare-related prediction tasks, with a focus on feature-driven disease detection.
Source Code