Closed Loop Paging Optimization for Efficient Mobility Management

The 4G/5G networks deploy conventional Tracking Area Update and multi-step paging procedures for mobility management. The paging procedure consumes significant amount of licensed spectrum resources. The signaling overhead is going to worsen further with the increasing use of small cells and higher user mobility speed. To address this challenge, the telecommunication industry is embracing closed loop approaches to predict user mobility patterns. Though the mobility pattern prediction is a known problem, most of the existing solutions apply it for enhancing the handover management and use academic dataset. Furthermore, limited research has been done on idle-state users. In this paper, we propose a Closed Loop Paging (CLOP) optimization solution using semi-supervised learning model to reduce paging overhead. We harness the real network dataset to analyze the location trail of users in the network to predict a subset of Base Stations for paging to locate idle-state users. Our empirical results demonstrate that Linear Support Vector Machine (L-SVM) classification method excels when compared to other supervised learning models. The L-SVM Classifier saves nearly 43% of the paging overhead with a marginal increase in connection establishment delay by around 7.3%.