Prediction of Parkinson’s Disease Using Machine Learning: A KNN-Based Classification Framework
DOI:
https://doi.org/10.63665/8rwsw211Keywords:
Parkinson’s Disease; K-Nearest Neighbors, Machine Learning; Feature Scaling; Random Over sampler; Predictive Diagnosis; Biomedical Voice Measurement; Flask DeploymentAbstract
Parkinson’s disease (PD) is a chronic, progressive neurodegenerative disorder characterized by the loss of
dopaminergic neurons in the substantia nigra, producing characteristic motor and non-motor symptoms in millions
of individuals worldwide. Early detection is clinically critical, yet traditional diagnosis is retrospective, subject to
inter-rater variability, and ill-suited to mass screening. This paper presents a machine learning–based classification
framework for PD prediction from biomedical voice measurements sourced from the UCI Parkinson’s dataset. The
proposed system employs a K-Nearest Neighbors (KNN) classifier augmented with systematic preprocessing,
MinMaxScaler feature normalization, and Random Over Sampler for class imbalance correction. Performance is
benchmarked against Logistic Regression, Support Vector Machine (SVM), Decision Tree, Random Forest, and XG
Boost. Hyperparameter tuning over k ∈ {1, 3, 5, 7, 9, 11, 13, 15, 20} identified the optimal KNN configuration. The
final model was serialized and deployed via a Flask web application, realizing an end-to-end, scalable clinical
decision support pipeline for early, automated Parkinson’s disease prediction
