神经网络在无线网络流量预测中的应用

发布时间：2018-03-09 09:31

本文选题：无线网络　切入点：流量预测　出处：《现代电子技术》2017年02期 　论文类型：期刊论文

【摘要】：无线互联网的流量数据非常分离且极其不稳定,混沌理论在其身上体现得特别明显,因此对无线网络流量进行预测具有一定难度。该文使用BP神经网络建立预测模型,在常规神经网络系统进行训练之前,需要对系统内部各个层次之间的连接权值以及阈值范围实行初始化操作,但是此操作将会影响神经网络最终收敛速度,有可能造成最终结果为非最优解,使得流量预测结果不是很理想。因此这里使用布谷鸟搜索优化方式对神经网络系统内各层之间链接值与阈值进行初始化操作,提高系统预测精度。该文使用遗传优化神经网络算法和粒子群优化神经网络算法建立同样的预测模型,并与该文研究的预测模型进行对比。实例分析结果表明,初期预测结果精度较高,与实际值比较吻合,但测试数据越靠后,预测值越不稳定,这主要是累计误差造成的。但总的来说,该文使用的布谷鸟优化BP神经网络预测模型的预测性能要优于由遗传算法和粒子群算法优化的BP神经网络。
[Abstract]:The traffic data of wireless Internet is very separate and unstable, and the chaos theory is very obvious on the wireless network, so it is difficult to predict the wireless network traffic. In this paper, BP neural network is used to build the prediction model. Before the conventional neural network system is trained, it is necessary to initialize the connection weights and threshold ranges between the different levels of the system, but this operation will affect the final convergence speed of the neural network. It is possible that the final result is non-optimal and the traffic prediction result is not very ideal. Therefore, the Cuckoo search optimization method is used to initialize the link values and thresholds between the layers of the neural network. In this paper, genetic optimization neural network algorithm and particle swarm optimization neural network algorithm are used to establish the same prediction model, and compared with the prediction model studied in this paper. The accuracy of the initial prediction results is higher, which is consistent with the actual values. But the more the test data, the more unstable the prediction value is, which is mainly caused by the accumulative error. But generally speaking, The prediction performance of the optimized BP neural network model used in this paper is better than that of the BP neural network optimized by genetic algorithm and particle swarm optimization.
【作者单位】：乌海职业技术学院;
【基金】：国家自然科学基金(23965184)
【分类号】：TN92;TP183;TP393.06

【相似文献】