Short-term bus travel time is an essential component of effective Intelligent Transportation Systems (ITS) including passenger information systems and transit signal priority (TSP). Several technologies such as Automated Vehicle Location (AVL) systems exist that can provide real-time information for bus travel time estimation. However, low resolution of data from such technologies presents a challenge in accurate travel time estimation. Several data-driven models for bus travel time estimation at signalized urban arterials are developed and tested. These models utilize low frequency AVL data, and only require knowledge of network specifications such as locations of bus stops and intersections. A linear regression model is developed which decompose total travel time into its components including free flow travel time, dwell time at bus stops, and delay at signalized intersections. A segment of Washington Street in Boston, Massachusetts, is selected as the study site. Various models are trained using Python libraries like scikit-learn and are then evaluated. The results indicate that the Support Vector Regression (SVR) model outperforms other regression models in terms of generalized error measures including mean absolute error (MAE), and root mean square error (RMSE). The findings of this study could lead to improved traveler information systems, TSP strategies, and overall contribute to better transit quality of service.