Research On Traffic Flow Prediction Based On Improved Whale Algorithm Optimization Of BiLSTM

In the context of the rapid development of urbanization in China,urban traffic management has been greatly challenged,and the establishment of ITS is an important way to solve this problem.Among them,traffic flow prediction is a key component of ITS,which can provide real-time,accurate and comprehensive traffic conditions for cities,which is of great significance for improving urban traffic conditions and alleviating urban traffic congestion.Due to the high randomness,ambiguity and nonlinearity of short-term traffic flow prediction,it also faces many problems and challenges.Based on this,this paper carries out the traffic flow prediction research based on adaptive Chirp modal decomposition and improved whale algorithm to optimize BiLSTM.The specific work is as follows:Firstly,aiming at the problem of uneven population distribution and slow convergence speed initialization of whale optimization algorithm,this paper adopts two improvement strategies.(1)In order to solve the problem of uneven population distribution initialized by whale optimization algorithm,it is proposed to replace the initial randomness value of WOA by using the Sobol sequence initialization method to improve the uniformity of population distribution and the global search ability of the population.(2)By adding segmented nonlinear convergence factor and adaptive weight,the influence of the algorithm on the position update of target prey is weakened in the initial stage,and the global optimization ability is strengthened.In the later stage,the influence on the position update of target prey is strengthened,and the late convergence speed of the algorithm is accelerated,thereby improving the optimization ability and speed of the algorithm.According to the above two strategies,the IWOA algorithm is constructed,and the algorithm is tested with six test functions.Secondly,the ACMD algorithm is used to decompose the traffic flow sequence and analyze its correlation.The advantages and disadvantages of EMD,VMD and ACMD data decomposition techniques were compared,and the traffic flow data sets were decomposed by ACMD decomposition techniques and their correlations were analyzed.Taking the traffic flow sequence as a random heartbeat signal,taking advantage of ACMD in processing multicomponent signals with overlapping frequencies,the traffic flow sequence is decomposed into four subsequences with limited bandwidth and their correlations are analyzed,and the strongly correlated sequence is selected as the research object.Finally,the IWOA-BiLSTM traffic flow prediction model is constructed.Adaptive Chirp modal decomposition was used to decompose historical traffic flow data,and four modal components were obtained and their correlations analyzed.The improved whale optimization algorithm was used to combine and optimize the number of input layer neurons,hidden layer neurons and learning rate of BiLSTM,and the target task was to minimize the root mean square error predicted by the BiLSTM model,and the hyperparameter optimization results of IWOA algorithm were introduced into the BiLSTM prediction model,so as to establish the IWOABiLSTM traffic flow prediction model.By comparing the prediction accuracy of each model,the forecast performance and accuracy of IWOA-BiLSTM were comprehensively evaluated.It is found that the IWOA-BiLSTM model has higher prediction accuracy than the other three models,and its prediction results are more consistent with the actual situation,which verifies the effectiveness of the improved whale algorithm and the superiority of the IWOA-BiLSTM prediction model.