Research On Method For Traffic Status Identification Of Highway Basic Sections

As a carrier of people's daily travel,highway is very important.As a result of the rapid increase of traffic volume,the scale of the road network is expanding,the structure of the road network is becoming more and more complex,and many other problems such as road traffic congestion tending to normalize,frequent road accidents,increasing energy consumption are becoming increasingly prominent.At present,the research results and practical application of expressway traffic condition discrimination have some foundation,but they are not very perfect.In addition,new data sources provide a new way to improve the method.Therefore,this paper carries out the research of traffic condition discrimination method based on microwave data and mobile phone data.Firstly,based on the basic section of expressway,this paper summarizes and analyses the commonly used freeway traffic flow parameter acquisition technology,and compares the advantages and disadvantages of various technologies,so as to clarify the data collected by microwave detector and mobile phone switching location detector as the data basis of the paper.In addition,the traffic parameters involved in the traffic state and their relationship are analyzed,and the traffic volume,speed and occupancy are selected as the traffic state parameters studied in this paper.Secondly,the basic principles of microwave detection technology and mobile phone handover positioning technology are elaborated,and the accuracy of microwave data and mobile phone data is analyzed.The traffic volume,microwave detection speed value,mobile phone detection speed value and occupancy rate are proposed as the parameters of traffic status discrimination research.Based on this,a traffic data preprocessing method is proposed,which focuses on the identification and repair of fault data and the transformation of traffic data.The respective characteristics of microwave data and mobile phone data are analyzed.The results show that the mobile phone speed measurement value is obviously higher than that of microwave speed measurement value,but the overall trend of change is consistent.Secondly,the criteria of traffic state partition are analyzed,and the traffic state partition is realized by the clustering analysis method of the fuzzy C-means algorithm.Aiming at the accuracy of the classification decision module in the fuzzy C-means algorithm,a classification decision algorithm based on random forest is proposed,and then a traffic status discrimination algorithm based on fuzzy C-means clustering and random forest is proposed.Aiming at the influence of unbalanced data sample set on traffic state discrimination,the method of synthesizing a few kinds of oversampling technology is used to balance the data and solve the problem of unbalanced distribution of actual data.A traffic state discrimination algorithm is established based on fuzzy C-means and improved stochastic forest.Then,taking the expressway as an example,the expressway state discrimination algorithm is analyzed and validated by an example.Firstly,this paper proves the feasibility of the random forest discriminant algorithm through two evaluation indexes,accuracy and recall rate.Then,the data balance is achieved by synthesizing a few kinds of over-sampling techniques.At the same time,the performance of random forest algorithm before and after improvement is compared.The results show that the improved random forest algorithm is relatively better.The performance of the improved stochastic forest algorithm is compared with that of gradient lifting algorithm,Adaboost algorithm and k-nearest neighbor algorithm.The results show that the discriminant algorithm based on fuzzy C-means and improved stochastic forest performs the best.The classification accuracy of training data and test data reaches more than 99%,while the running time is the shortest.Finally,based on the actual data of highway,the highway status identification method based on fuzzy C-means and improved stochastic forest algorithm is analyzed and validated.Firstly,this paper proves the feasibility of the random forest discriminant algorithm through two evaluation indicators: accuracy and recall rate.Then,the data balance is achieved by synthesizing a few kinds of over-sampling techniques.At the same time,the performance of random forest algorithm before and after improvement is compared.The results show that the improved random forest algorithm is relatively better.Then,the performance of the improved stochastic forest algorithm is compared with that of gradient lifting algorithm,Adaboost algorithm and k-nearest neighbor algorithm.The results show that the performance of the improved stochastic forest algorithm based on fuzzy C-means is the best.