The 5th generation (5G) mobile network has been deployed globally for many years. Compared with Long-Term Evolution (LTE), there are several major improvements, such as throughput, latency, and energy efficiency. The energy efficiency of user equipment (UE) is one of the critical concerns and has been discussed and specified in different updated releases by the 3rd Generation Partnership Project (3GPP) to minimize the UE consumption power. Therefore, it is necessary to understand the behavior of UE power consumption in different network configurations.
In this thesis, the power consumption models of an UE, specifically a millimeter wave (mmWave) UE, are examined. The study was performed in a non-standalone (NSA), where both Long-Term Evolution (LTE) and 5G are connected for dual connectivity. Three modeling methods are investigated: polynomial regression, decision tree regression, and neural network. The results show that the decision tree regression and the neural network are versatile in different test schemes and perform well in prediction, while polynomial regression performs well only with power sweep, i.e., the simple curve of the target variable. Moreover, adding a few highly correlated features improves the performance of the model.