KEY PERFORMANCE INDICATORS OPTIMIZATION FOR LONG-TERM EVOLUTION NETWORKS USING MACHINE LEARNING

Abstract

 Some COPs (Configuration and Optimization Parameters) that may be manually changed to manage the network and deliver improved QoS (Quality of Service) are a problem for the present LTE community. In the future, the range of optimization parameters is predicted to reach 2000 per BTS. Fine-tuning these KPI parameters manually is not an easy task, aside from those thousands of KPIs rising in networks which will increase the weight of the Service providers in optimizing and managing the network. Subsequently, we suggest a model framework blended with a new method including Machine Learning to find out the premier combination of two parameters, HOM (Handover Margin) and CIO (Cell Individual Offset). The first part of the framework leverages to know how to forecast the KPI of the Network given numerous exceptional mixtures of HOM and CIO. The obtained estimations are then sent into a GA (Genetic Algorithm) that searches for the two parameters' ideal combinations to offer the highest SINR for all users.