The purpose of the study was to investigate the use of machine learning (ML) in diagnostic measures of essential tremors (ET) versus tremors associated with Parkinson’s Disease (PD). It is important to distinguish these phenotypically similar tremors so that effective treatment can be employed. Using an accelerometer, raw data from 226 PD diagnosed and 188 ET diagnosed patients were collected. Supervised ML algorithms, including support vector machines (SVM), were used to classify the data. The models were trained and tested using 10-fold cross validation to ensure accuracy. Traditional diagnostic methods such as the Tumor Severity Index (TSI) provided a 70.5% accuracy compared to the 81.1% accuracy of ML classification using triaxial data for both. However, the variability in data collection methods across the six different centers involved in the study likely affected the model’s performance, as there were relevant differences among the cohorts from each center.