Research On Traffic Flow Forecast Based On Convolutional Bidirectional Long Short Term Memory Network

Nowadays,traffic demands are growing increasingly,followed by a variety of traffic problems.What need to be solved immediately is the traffic congestion.In order to alleviate and manage the traffic congestion,short-term traffic flow forecasting is necessary for decision-making by the traffic control department.Accurate short-term traffic flow prediction can provide real-time road condition information for urban residents and improve the traffic capacity of urban sections.With the emergence of the era of big data and the continuous enhancement of information technology,the resulting data is getting more and more enormous,especially traffic data.Therefore,how to use these massive traffic data to predict the future traffic flow is a hot spot in this area.Against the background,this paper presents the shortcomings of the current short-term traffic flow prediction models,and proposes a model combining convolutional neural networks and bidirectional long short term memory networks to extract the spatial and temporal characteristics of short-term traffic flows to predict flow.Based on this,this paper has completed the following works:To begin with,the paper systematically elaborates the existing research results of current traffic flow forecasting;secondly,it introduces the relevant theories,model parameters,and correlation analysis of traffic flow,and analyzes the relevant methods and models of current short-term traffic flow prediction.Furthermore,it points out the shortcomings of the trajectory data and the method of obtaini-ng traffic flow.Finally,a spatio-temporal prediction model based on Conv-BiLST(Convolutional Bi-directional Long Short Term Memory Networks)is established to perform long-term and short-term traffic flow prediction.This paper uses Kunming taxi trajectory data to verify the Conv-BiLSTM model.Because the amount of data is too large,this paper proposes a matching algorithm based on Storm which greatly saves preprocessing time.By integrating the short-term traffic flow prediction and the accuracy rate reaches 93.8%,which proves that the accuracy and effectiveness of Conv-BiLSTM model.It can more effectively reflect the characteristics of the spatiotemporal data.