摘要
针对BP神经网络算法存在容易陷入局部极值、收敛速度慢、寻优精度低等问题,采用改进的PSO算法和GA算法对BP神经网络算法进行优化。对PSO算法中惯性权重公式进行改进,重新调整速度更新公式,提高算法稳定性。优化后期,引入GA算法的交叉、变异操作来扩大粒子的搜索空间以提高粒子的多样化,避免粒子过早收敛到局部解,从而提高BP神经网络算法的精度。采用改进后的BP神经网络模型对移动用户行为进行预测。实验结果表明,改进后的网络模型可以有效地提高移动用户行为预测的准确率和实效性,发掘出用户的行为特征和使用业务的规律,在一定程度上可为移动通信网络的质量优化和市场运营提供理论依据,帮助网络运营商提升网络服务水平。
In viewof the problem that BP neural network algorithm has lowspeed of convergence,lowprecision and easily falling into the local extremum,we optimize BP neural network algorithm with improved PSO algorithm and GA algorithm.The stability of the algorithm can be enhanced by improving inertia weight formula and adjusting speed updating formula in the PSO algorithm.In the later stage of optimization,adopting the crossover operation and mutation operation of GA algorithm can expand the search space of the particles,which can improve the diversification of the particles and prevent particles from falling into the local solution previously and improve the accuracy of the BP neural network algorithm finally.We adopt the trained BP neural network model to predict the behavior of mobile users. Experiment shows that the optimized network model can improve the accuracy and effectiveness of behavior prediction of mobile user effectively and discover the behavior characteristics and rules of users business,which can provide theoretical basis to the quality optimization of mobile communication network and market operation and help network operators to enhance the level of network services.
作者
陈春玲
陈红
余瀚
CHEN Chun-ling;CHEN Hong;YU Han(School of Computer,Nanjing University of Posts and Telecommunications,Nanjing 210003,China)
出处
《计算机技术与发展》
2018年第7期178-181,186,共5页
Computer Technology and Development
基金
国家自然科学基金(11501320)
关键词
BP神经网络算法
PSO算法
GA算法
行为预测
BP neural network algorithm
particle swarm optimization
genetic algorithm
behavior prediction
