As lithium-ion batteries (LIBs) become increasingly more common, accurate and efficient methods must be developed to predict their state of health (SOH). Battery management systems (BMS) could be greatly improved with an accurate prediction of SOH, which would result in better battery performance and ensure safety. While using advanced electrochemical modeling and simulations to predict SOH is accurate, the models are extremely complex, and significant computational time is required. Machine learning has proven to decrease the time needed to compute and maintain the accuracy of predictions. In this work, XGBoost, which has the advantage of fast computation time and scalability, is utilized as the machine learning algorithm to predict SOH using dataset from a physics-based battery degradation model. Datasets are generated using dynamic cycling protocols to consider the realistic fluctuating charge and discharge cycles in real transport and stationary applications. The developed method can predict SOH using multiple combinations of degradation parameters as input. Accumulated charge and discharge, state of charge, and applied current are found to be important datasets to accurately predict SOH. The proposed method can predict the SOH of batteries in a significantly shorter time than current electrochemical models while maintaining the high accuracy required for a BMS.