Research And Application Of Highway Condition Prediction Based On Big Data Of Vehicle Traffic

With the continuous construction and improvement of the highway network,a large number of monitoring devices are used at highway toll gates and road centers to record passing vehicle information,thus massive vehicle traffic data is stored in the highway monitoring system.How to effectively dig out more valuable information from these popular data has become one of the research focuses in recent years.In this master thesis,we will combine data mining technology and big data technology,through the mining and analysis of massive traffic data,get the traffic volume and traffic speed information of highway sections,so as to realize the prediction of future traffic flow and traffic speed.On this basis,by using the predicted traffic flow and traffic speed and other factors to build a highway traffic condition evaluation system,so as to predict the future road conditions of the highway,provide a valuable reference for highway operation management and residents’ travel.The main work and innovations of this master thesis are as follows:1.In view of the problem that the traditional traffic flow prediction model only considers the temporal characteristics of the traffic flow and leads to poor prediction results,this master thesis combines the temporal and spatial characteristics of the traffic flow and proposes a convolutional neural network(CNN)Traffic flow prediction model combined with LightGBM.The model first uses the CNN model to mine the temporal and spatial correlations between the monitoring points and entrances and exits of adjacent road sections of the expressway to realize the extraction of spatiotemporal features of traffic flow data,and then inputs the extracted feature vectors into the LightGBM model to achieve Stream prediction.Experimental results show that the model has better prediction performance than previous models.2.In terms of traffic speed prediction,considering the limitations of Long Short-Term Memory(LSTM)network models that can only extract traffic speedfeatures in one direction,the traffic speed prediction model of Bidirectional Long Short-Term Memory(BiLSTM)network based on attention mechanism is designed.The model first uses the BiLSTM model with the combination of forward LSTM and reverse LSTM to extract traffic speed features,and then introduces an attention mechanism to highlight the features that play a key role in the outcome of traffic speed prediction,and assigns them different weights.The experimental results show that the traffic speed prediction model based on the attention mechanism of BiLSTM network can improve the prediction accuracy.3.According to the predicted traffic volume and traffic speed,an improved fuzzy comprehensive evaluation method is proposed for evaluating the highway condition.The algorithm combines the highway traffic conditions of the expressway,converts the predicted traffic flow and traffic speed into the future roadway saturation,traffic speed and traffic density and other highway roadway condition evaluation indicators,and uses fuzzy comprehensive evaluation method to calculate the future traffic condition grade of highway section.In terms of weight coefficient calculation,this master considers the influence of the two situations of human subjectivity and objective facts on the experiment,and also considers the problem of uncertainty in expert scoring,a method based on the combination of Interval Analytic Hierarchy Process(IAHP)and entropy weight method is proposed to improve the weight coefficient.By using this method,the weight coefficients of day and night are calculated respectively,and then the membership of each index at different times is determined by the trapezoid membership function,and finally the highway conditions at different times are obtained.After comparing with the actual situation,it is found that the judgment method conforms to the changing law of highway conditions.Finally,combined with the actual project,the above research results were used to design and develop a highway traffic condition prediction platform to achieve a visual display of the research results.The platform integrates the functions of data collection and query of passing vehicles,query of highway traffic conditions,query of highway traffic conditions,and query of highway traffic conditions.The system can show the future road conditions,and can better meet the needs of practical applications.