Parkinson’s disease (PD) is a gradually progressive disease that affects the nervous system and causes difficulty in movement and other disabling symptoms. This research investigates the identification and severity forecasting of Parkinson’s Disease (PD) utilising voice samples. The MSR dataset, which includes voice features from healthy individuals and patients with Parkinson’s disease was analysed to determine significant differences in vocal characteristics. Key features, including jitter, shimmer, and harmonicity, were extracted to distinguish between affected and non-affected individuals. Various machine learning models,CNN, XGBoost, LightGBM, and Random Forest, were utilised to classify the presence of PD and to predict scores on the Unified Parkinson’s Disease Rating Scale (UPDRS), thereby evaluating disease severity. This study additionally aimed to predict motor and total UPDRS scores utilising a distinct telemonitoring dataset. Advanced regression techniques were utilised to enhance the precision of severity predictions. The analysis of real-world voice samples using Praat software improved the feature extraction process, providing greater insights into vocal changes associated with PD. The initiative seeks to enhance patient outcomes and support clinicians in more effective disease management through timely interventions.