Research On Traffic Prediction Of Urban Road Traffic Flow Data And Microblog Related To Traffic

With the improvement of urban road traffic system and the improvement of people's living standards,the number of motor vehicles has been increasing,and the traffic congestion has become a “urban disease” in the worldwide.Road traffic prediction is of great significance to urban traffic control system and traffic management system.Accurate traffic prediction can provide travel guidance for travelers,avoid traffic jam and improve the overall efficiency of the urban road network.In addition,with the rapid development of the Internet,people tend to share what they have seen and heard on social networking platforms.The text not only constructs a traffic prediction model for urban road traffic flow data,but also extracts and analyzes the text related to traffic events on the microblog to achieve the purpose of traffic prediction.In constructing the traffic prediction model for urban road traffic flow data,the traffic flow data is firstly denoised and analyzed.Aiming at the time-varying and non-linearity of traffic flow data,BP neural network model is selected as the basis of prediction model.It is proposed to improve BP neural network prediction model by self-mutation fireworks algorithm,because BP neural network has some shortcomings such as easy falling into local minimum.The dynamic optimization of BP neural network algorithm parameters avoids the problem of over-fitting and falling into local minimum in the optimization process.The prediction experiments were carried out on multiple sections of the Beijing Second Ring Expressway in multiple time periods.The relevant indicators prove that the model proposed in this paper is more suitable for short-term traffic flow prediction.To construct the traffic prediction model based on traffic text on the microblog,firstly,the text related to traffic events is obtained by using web crawler technology.In view of the simplicity and randomness of language on the microblog,a multi-feature word function screening model and a text representation model are constructed to extract and express feature words from traffic texts on the microblog.Traffic events are identified by a combined classification method of Bayesian classifier,K-nearest neighbor classifier and support vector machine classifier.Also traffic information will be extracted to realize road traffic prediction.The self-variation fireworks algorithm improves the BP neural network prediction model to describe the future situation of road traffic by numbers,but the model fails to predict the sudden change of traffic flow data,and it is impossible to give a clear conclusion when the predicted value is small.The traffic prediction model based on the traffic text from the microblog can solve these two kinds of problems well,but the model is limited by the microblog published by bloggers,and it is impossible to predict the road traffic of any road section at any time.Therefore,the method of integrating two kinds of prediction models can accurately predict road traffic.