In this paper, the main aims are to obtain a quality of signal prediction model in idle-static mode and a quality of connection prediction model in connected-mobile mode, for this, we apply machine learning technique to a real dataset collected from a LTE network deployed at Quito, Ecuador. The proposed models are capable to predict the conditions of low received signal strength and low data rate which is important to select the appropriate method that will most likely offer the highest quality of service. The proposed schemes based on decision tree improves in the idle-static mode and present an accuracy of 67% approximately when compared to the connected-mobile mode, and finally, we propose future works.