Machine Learning based Prediction of Parkinson's Diseases

Parkinson's disease(PDs) is a neurological disorder characterized by accidental or uncontrollable movements such as shaking, rigidity, and trouble with balance and coordination. Symptoms are usually minor at first and intensify over time. PD datasets tend to be compact and inadequately diversified. Several datasets contain incorrect values or discrepancies as a result of diagnostic procedure inconsistencies. Datasets frequently have fewer PD patients than healthy individuals, resulting in biased models. SVM is a common method for PD prediction due to its ability to handle tiny, high-dimensional data sets and its robust classification skills. SVM performs effectively even with small medical datasets, unlike deep learning models, which require enormous volumes of data. SVM minimizes overfitting, particularly when working with tiny datasets, making it more dependable for use in real-world medical settings. The study employed SVM algorithms to predict Parkinson's disease. Additionally, disease data is classified and compared using SVM with different kernel parameters. The study compares the performance of four alternative SVM models Linear SVM (LSVM), Quadratic SVM (QSVM), Cubic Gaussian SVM (CGSVM), and Medium Gaussian SVM (MGSVM) using five assessment metrics: accuracy, sensitivity, specificity, precision, and F1-Score. MGSVM is the most effective model for predicting PDs, with the highest precision and consistency.