Research On Traffic Forecast Based On Reinforcement Learning And Evolutionary Algorithm Optimization

In the field of network monitoring and operation and maintenance.How to reasonably use historical traffic data to predict traffic changes in a period of time in the future has very meaningful socio-economic value.In recent years,time series models based on deep learning technology have been increasingly applied to this field.Deep learning models are highly dependent on data preprocessing and various hyperparameter adjustments in the model training process.The traditional way is to ask some experts with field and artificial intelligence background to give empirical values.This method has high cost and difficult promotion.At the same time,manual experience lacks certain accuracy and interpretability.Therefore,the introduction of automatic optimization parameter search algorithm in parameter and hyperparameter adjustment has urgent research needs and socio-economic value.Aiming at the difficulties of automatic parameter optimization encountered in the development of network traffic forecasting systems,this paper,combined with recent research results of enhanced learning and evolutionary computing,A set of schemes based on improved Q-Learning strategy and Levy's flight combined with lightning optimization algorithm are proposed.Automatically search for optimal parameters in the data preprocessing stage of network traffic prediction and the deep learning model training stage.The research content and main work of this paper are as follows:Analyzing the characteristics of the network traffic data set,the key index vacancy rate parameters in the process of cleaning the data set.This paper proposes a fast estimation network model based on process compression and an optimized parameter search algorithm for Q-Learning(QV-QL).The model starts from a predictive model based on deep learning,and on the basis of ensuring the functionality and certain accuracy of the model.Through the compression process and the introduction of mixed-precision calculations,the speed of searching for optimal parameters has been greatly improved.An optimization algorithm for the lightning attachment process based on Levy flight improvement(Levy-LAPO)is proposed.Through the overall driving ability of Levy's flight,solved the problem of slow convergence.This paper compares the improved algorithm with the classic algorithm on standard functions and real data sets to verify the superiority of the improved algorithm.Combining the above two research contents,a network traffic prediction platform based on the SEQ2 SEQ model is designed and implemented.And verified the proposed algorithm in the system.Experiments show that the optimization parameter search algorithm proposed in this paper can find better parameters and hyperparameter selections than manual experience within a reasonable running time.Realize the end-to-end automatic optimization process without domain knowledge intervention.