Parkinson’s disease (PD) is a neurodegenerative disorder characterized by motor or non-motor symptoms, which are typically assessed through subjective clinical evaluations. Based on physician expertise and patient-reported symptoms, these assessments can lead to inconsistencies. This study aims to provide an objective and standardized method for quantifying tremor severity using wearable sensor technology and machine learning (ML). A total of n = 58 PD patients was collected using a nine-axis Inertial Measurement Unit (IMU) to measure hand tremors in both the postural and rest states for left and right hands. Data were analyzed by extracting features from the statistical, time, and frequency domains, which were then used to classify tremor severity into four categories: slight, mild, moderate, and severe. To enhance model performance, various machine learning models, including XGBoost, Random Forest, SVM, and KNN, were evaluated with hyperparameter optimization and 5-fold cross-validation. The XGBoost model achieved the highest accuracy for postural and rest tremors. Our results demonstrate that wearable sensor data combined with machine learning enables a reliable, interpretable, objective assessment of Parkinsonian tremor severity, offering clinicians a tool to improve decision-making and personalized disease management.