Handover Analysis of Mobile Cellular Network in a Populated City

Mobile network operators continuously monitor QoS in terms of multiple metrics to ensure high network performance. Handover (HO) is one of the key issues in cellular communication networks, as it poses multiple threats to QoS. In this work, several real measured radio frequency (RF) parameters are used to analyze handover performance using statistics and Machine Learning. The measurements were collected through Android applications in 4G mobile networks in an urban area of the city of Quito, Ecuador where the existence of nearby base stations was checked. Variables were created for the analysis, such as one indicating the existence of a handover and another indicating whether the handover was successful or unsuccessful. The results of the statistical analysis show that performance metrics determine HO problems. While Machine Learning was used to classify whether the handover was successful or unsuccessful based on certain input parameters using Decision Trees. A predictive analysis was also made using Linear Regression to determine whether handover existed or not. Finally, the machine learning models were evaluated, obtaining an accuracy of 0.73 for the decision tree and an RSME of 0.077 for the regression.