Application Study Of Network Traffic Prediction Based On Optimized Neural Network

With the continuous development of the global information intelligence process,the scale of the network is multiplying every day.As problems such as changes of the requirements of various of application emerge in endlessly,the manual maintenance of the network is no longer in step with the times.Based on a philosophic idea of nipping in the bud.Using scientific and technological methods to predict future network traffic is gradually becoming a research hotpot.With the rapid development of computer technology in recent years,Deep Learning has made breakthrough progress in Data Mining,Voice Recognition,Natural Language Processing and other fields,which will bring new opportunities to network traffic prediction.Based on the Seq2 Seq model,this thesis builds a network traffic prediction model,which will be applied into a network traffic prediction project for a city in Hubei province.The main work is as follows:1.Analyze the advantages and disadvantages of several classic Neural Networks,and select out a suitable Neural Network as this research topic of the thesis,The selected Neural Network will be used as the theoretical basis of this thesis to construct the network traffic prediction model in this thesis.2.Optimize the initial Learning Rate of the network data flow prediction model.By testing Single-peak and Multi-peak function of the optimization algorithms that participated in the comparison the overall analysis of the experimental results shows that the Dragonfly Optimization Algorithm has excellent global search and local development capabilities during independent optimization,which provides necessary conditions for optimizing the network model Hyperparameters.Through the analysis of the experimental results,the convergence time of the loss function of the network model optimized by the Dragonfly Optimization Algorithm becomes shorter,thereby accelerating the training speed of the network model.3.Optimise the batch size of the network data flow prediction model.We regard different sizes of batch size value as the Hyperparameters of the network model input to the network model,participate in the training of the network model,and analyze the batch size of the appropriate size from the experimental results can speed up the training speed of the network model,which proves that the batch size will affect the optimization of the network model and the speed of model training.We applied the constructed network traffic prediction model to a city in Hubei province.From the experimental results,it is analyzed that the prediction accuracy of the network model reaches the expected prediction accuracy.In general,this thesis constructed the optimization of the network traffic prediction model and network model Hyperparameters based on the Seq2 Seq theory in the Recurrent Neural Network.The corresponding experimental results prove that optimization method has improved the training speed of the model and the construction of network model achieves the expected prediction accuracy in the network traffic prediction,it has high research value and broad application scenarios.