Electric Submersible Pump (ESP) is an effective artificial lift method of pumping medium to high production fluids. However, it is prone to failure because most of the system is located downhole. When mechanical defects start to develop, vibration signal changes. By examining the variation in vibration, component failures can be located and predicted. The objective of this study is to creates a machine learning model to predict ESP vibration using experimental data. The dataset that was utilized to create the model has more than a million data points collected from laboratory. Based on pump performance, the experimental data are divided into three groups: normal, advised and failing pump conditions. The model features are operating time, pump speed, oil flow rate, water flow rate, pump intake pressure, motor temperature, free gas percentage and liquid/mixture density and viscosity. Among the algorithms used to develop the model, random forests regression produces results with acceptable accuracy while not requiring too much computing power. Because of that, random forests regression was used for further optimization and simulation. The results of the model show that pump speed and operating time have the biggest effects of all the features. Due to the limited range of experimental data, in some cases, it is either unknown or unclear how the features affect ESP vibration, particularly in advised and failing pump conditions. The model's vibration predictions closely match the results of the experiments. This model can be used to forecast pump vibration under a particular operating condition by incorporating into the ESPSim program. The predicted vibration magnitude can be compared to the actual value to determine whether an ESP is in failing condition. With this information, the amount of downtime when the pump fails could be minimized. The extreme downhole conditions can also cause sensor cables to break and become incapable of transmitting data to the surface. Operators can use predicted ESP vibration to get an idea of the status of the pump.