The control of power systems can be enhanced through wide area monitoring, where measurements from devices such as phasor measurement units are utilized to enhance the system awareness. These measurements can be utilized for dynamic state estimation, where the rotor angle and speed of the synchronous machine are estimated. This paper proposes a new scheme which utilizes the unscented Kalman filter (UKF) along with machine learning for dynamic state estimation. The machine learning algorithms are used as a support system for enhancing the robustness and quality of the UKF and the overall estimates. The new scheme is tested on the Western System Coordinating Council 3-machine 9-bus test system under different operating conditions.