In power systems, there are two different types of oscillation mechanisms–natural and forced oscillations. Natural oscillations are caused by system disturbance when the damping of the system is not sufficient. Forced oscillations are typically caused by external input driving the system into a sustained oscillation. Distinguishing low-frequency oscillations is a prerequisite to dealing with oscillation events in power systems. This paper attempts to distinguish between natural oscillations and forced oscillations using machine learning technologies. Decision tree, support vector machine, neural network, and convolutional neural network algorithms are evaluated. Transfer learning is applied to overcome the lack of training data.