基于改进型深度学习的流量预测

为了解决无线网络中流量的预测精度不高的问题，提出了一种自适应分组的栈式自编码(AG-SAEs)深度学习预测方法。在数据的预处理过程中，首先使用最大最小方式对数据进行归一化处理，并提出一种新型的自适应分组方法，把归一化后的链路数据进行关联性分组；然后，基于深度学习方法建立了一个多输入多输出的预测模型，并将分组后的数据输入到预测模型中，对该模型进行训练来建立输入和输出流量之间的映射关系；最后，为了进一步提高预测精度，在模型的训练过程中，使用改进型的牛顿法来进行权值参数更新。仿真实验以及和其他算法对比的结果证实了所提方案具有更小的预测相对误差。

Traffic prediction based on modified deep learning

To solve the problem of low traffic prediction accuracy in wireless networks,a method based on the adaptive grouping stacked auto-encoders(AG-SAEs) deep learning is proposed.In data preprocessing,the maximum and minimum method is used to normalize the data,and a novel adaptive grouping method is adopted to divide the normalized data into different groups adaptively. Then,a multi-input multi-output prediction model based on the deep learning model is established. All the groups are input to the stacked auto-encoder model to train the model and map the relationship between input and output traffic. Finally,in order to further improve the prediction accuracy,the modified Newton method is used to update the weight parameters in the model training section.The simulation experiment and comparison with other methods show that the proposed method processes a smaller prediction relative error.