Research On Detection And Correction Algorithm Based On Section Traffic Anomaly Data

At present,large and medium-sized cities in China,even some small cities,have their own traffic database,which contains great potential data value.Unfortunately,however,these huge and complicated traffic data are not fully utilized.Among them,there are many reasons.The most important one is that the reliability of some original data collected by traffic monitoring equipment is low,which directly results in the impact and even misjudgment of subsequent data mining based on this.Therefore,it is very important to detect and correct the anomaly of traffic original data.In this article,we will study the section traffic anomaly detection data correction algorithm,through the establishment of two algorithms of different traffic data quality control process,so as to improve the accuracy of traffic data,for subsequent transport planning,traffic control,and even across social application in the field of providing quality data to support.Firstly,the research status at home and abroad is analyzed and summarized from the two aspects of anomaly data detection and correction.In general,relevant algorithms are based on traditional mathematical statistics and deep learning based on neural network,so these two directions are determined as the algorithm ideas,and the main research content and technical roadmap of this paper are proposed at the same time.Secondly,through the analysis of the three-parameter characteristics based on the original traffic data,the correlation of each parameter data is obtained,providing theoretical basis for the subsequent algorithm.Then,through the analysis of the abnormal data set,the overall impact on the data quality was obtained,and the K2077 section data of Shantou-Kunming expressway was used for example verification.Then,two algorithms based on traditional mathematical statistics and deep learning based on neural network are used to detect and correct cross section traffic data.According to the cross-sectional data of ShantouKunming expressway K2077,multiple gaussian detection and GM(1,1)-Markov correction of volatility optimization were used to control the quality of the original data set from the perspective of traditional mathematical statistics.A detection and correction algorithm based on optimized recursive network LSTM(long-term short-term Memory)was used to control the quality of the original data set from the perspective of deep learning based on neural network.The first algorithm is programmed on matlab,and the second algorithm is programmed on jupyter platform based on tensorflow framework.Finally,the two algorithms mentioned in this paper are compared with the common traffic data preprocessing algorithms.The results show that the performance of the two algorithms is better than the general traffic data detection and correction methods.At the same time,the two algorithms have relative advantages in different application backgrounds.