Wireless Network Parameters Optimization Method Based on Behavioral Learning

To solve the problem of customer complaints caused by prolonged optimization cycles for Wireless Network Performance Degradation cells, a method based on multi-objective reinforcement learning for network parameter optimization operational behavioral learning is proposed. Firstly, behavioral data of experts engaged in network parameter optimization are collected and annotated with states. Subsequently, a comprehensive multi-objective network optimization model is developed that takes into account key performance indicators of network issue cells, other performance metrics, and the improvement levels of neighboring area performance indicators. Lastly, a multi-objective reinforcement learning approach is formulated using the TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution) method to construct a multi-objective reward function. This approach is combined with proximal policy optimization algorithms to realize the recommendation of wireless network optimization strategies. The goal is to enhance the efficiency of network parameter optimization and improve user perception of wireless network performance. To validate the feasibility and superiority of this approach, a comparative analysis is conducted against various alternative methods, including those based on expert experience, association rules, decision trees, and other deep reinforcement learning techniques such as Deterministic Policy Gradient (DDPG) and Deep Q-network (DQN). These methods are utilized to independently perform parameter optimization for the performance indicators of 1000 network issue cells. The results indicate that the parameter optimization scheme of this study achieves higher network issue resolution rates and greater improvement in cell performance indicators compared to other methods.