Repair Of Low Rank Matrix Traffic Flow Data Based On Neural Network

With the development of information technology,transportation data has entered the era of big data.At the same time,big data contains a lot of important information,especially the traffic flow data,and its changes reflect the state of transportation,such as The situation of road congestion can be seen,so the observation and research of traffic flow data is conducive to solving traffic road problems.However,traffic data is usually due to various problems in the process of collection,transmission or storage,and the data will be lost and wrong.The lack of data not only reduces the validity of the data,but also brings great obstacles to the study of traffic conditions.Repairing missing traffic flow data can deepen the understanding of the spatiotemporal correlation and statistical characteristics of traffic flow data on the one hand,and have important significance for the subsequent analysis of intelligent traffic systems.Therefore,the research on the restoration of missing traffic flow data has important theoretical and practical value.The research goal of this paper is to design traffic flow data repair algorithms with high accuracy.The main research contents and research results of this article are as follows:1.This paper first explains in detail the collection methods of traffic flow data and the sources of missing data,and then analyzes the characteristics of the traffic flow data.Then,through research,it is found that the low-rank matrix repair model is suitable for the repair of traffic flow data.There are three types of data missing modes: completely random missing(MCAR),random missing(MAR),and mixed missing(MIXED),so the experiment needs to preprocess the missing data on the original data,and then use Matlab on the processed data.The implemented model performs data repair on the data of the three missing modes,and the experimental results can be seen through comparison.2.Establish a traffic flow data prediction model,and use the currently popular neural network LSTM model to repair the three missing modes of traffic flow.The LSTM model is written using the keras framework and Python code.Part of the pre-processed data is input into the built LSTM network,LSTM determines the network parameters and weights by learning the characteristics of the data,and outputs traffic flow data for the next day.It is found through experiments that the LSTM model repairs the three data loss modes effect.3.Because a single data repair method is often difficult to achieve accurate repair,this chapter proposes an LSTM model integrated repair model based on a low-rank matrix completion model and deep learning.Two models are used to repair the mixed missing mode,and finally the integrated model is obtained.Forecast results.