Closed Loop Optimization (CLO) is a solution to enhance mobile network performances based on network counters using reinforcement learning and supervised machine learning algorithms. MOON-CLO automatically detects a mobile network’s performance problems and their root causes when performance target(s) are provided.
MOON’s CLO feature comprises of two parties – CLO modeling and CLO manager. CLO modeling models the target performance to improvement in terms of the imported network KPIs and counters so that the problematic cells which do not meet the target performance level and root causes of low performance. The expected improvement gain is also assessed within the CLO modeling for individual problematic cells.
CLO manager selects the proper solutions to improve the target performances of the problematic cells corresponding to the identified root causes and the assessed improvement gain. When the effect of the proposed solution are monitored and its improvement is judged automatically after the proposed solution has been actually implemented into the mobile network. Based upon the effectiveness of solutions, CLO manager will continually apply the same solution with different aggressiveness of parameter values or different solutions based on reinforcement learning algorithm. The solution effectiveness and the problematic cell’s KPI, counter traits are also trained using supervised machine learning algorithm to choose the target cells which show higher probability of performance improvement when new target area is provided. Ensemble machine learning by combining various cell classification algorithms is used to classify the cells to improve the accuracy of cell classification.
Corresponding to the identified root causes and low performance and the cell classification, different improvement solutions are recommended to individual cells without human intervention.
MOON-CLO system employing cutting-edge AI technology can be used to solve various optimization problems not only for mobile network performance optimization.