OPTIMIZATION AND EVALUATION OF SUPPORT VECTOR MACHINE PERFORMANCE FOR PARKINSON’S DISEASE CLASSIFICATION USING BIOMEDICAL DATA

Satriyo Kristanto (Health Information Management, Kediri Baptist Hospital Institute of Health Sciences)
Syndia Puspitasari (Health Information Management, Kediri Baptist Hospital Institute of Health Sciences)
Muhamad Liswansyah Pratama (Data Science, East Java Veteran National Development University)

JTH: Journal of Technology and Health
Vol. 4 No. 1 (2026): July: JTH: Journal of Technology and Health

Dysphonia, hypophonia, and decreased pitch variation are common vocal deficits linked to Parkinson's disease, a progressive neurological illness. These anomalies can be used as non-invasive biomarkers to identify diseases early. Using biomedical speech data, this work attempts to optimize and assess the Support Vector Machine (SVM) algorithm's performance for Parkinson's disease categorization. The UCI Machine Learning Repository provided the dataset, which included 195 observations with 23 biological voice variables, such as jitter, shimmer, and characteristics related to frequency. Cleaning, standardization, and dividing the dataset into 80% training and 20% testing sets were all part of the data preprocessing procedure. To find the best SVM configuration, Grid Search and k-fold cross-validation were used for hyperparameter tuning. The findings of the experiment demonstrated that the Radial Basis Function (RBF) kernel.