Research On Short-term Flow Forecasting Based On Common Traffic Flow Seeking Under Big Data Environment

Forecasting the traffic flow is one of the significant researches of intelligent transportation based on traffic data.In July 2016,implementation of the national development and Reform Commission,the Ministry of transport issued "To promote the implementation of the program" Internet plus "convenient transportation to promote the development of intelligent transportation",and proposed the construction of the advanced awareness monitoring system,the construction of the next generation of traffic information basic network,and strengthening the opening and sharing of transportation information and other specific requirements.The combination of big data and traffic is considered to be an effective means to alleviate traffic congestion,reduce traffic accidents and improve the efficiency of traffic.Short-term traffic flow forecasting is the core part of intelligent traffic control.It is the precondition of precise traffic control,effective traffic guidance and accurate information service.The research of short-term traffic flow forecasting has a long history.With the development of the times and the progress of technology,the research method is also updating.In the past,most of the prediction methods were calculated mainly by pure mathematical model,and mathematical inferences were carried out by mathematical statistics model,and the law of traffic situation evolution was ignored.In this paper,a short term traffic forecasting based on fusion distance of similar traffic flows seeking model is proposed.The socalled similar traffic patterns seeking prediction means the method of predicting the future flow by looking for historical similar traffic situation in big data environment.On the basis of traffic data analysis,this paper studies the similar traffic flow phenomenon in traffic.Based on this,a model for short-term flow forecasting is built;using hierarchical clustering method to construct the database of historical traffic flow proposing a new method of fusion distance calculation combined with Euclidean distance and cosine distance;exploring the influence of the key parameters in the model on the prediction effect,and the value of the parameters is given.The specific contents are as follows:(1)Data processing and analysis.The original data from multiple sections of RFID and video clip flow data are analyzed and processed.There are many problem data in the original collection data,and the processing methods for various data problems are studied,and the processing effect of the actual data is given.(2)Research on traffic situation characteristics.The analysis of the characteristic traffic flow under various traffic conditions is carried out to explore the influence of these factors on traffic flow and the characteristics presented.This paper analyzes and explores the feasibility of mining data with similar change situation from mass data to predict short-term traffic volume,and puts forward the corresponding short-term traffic prediction algorithm.(3)Section database clustering research.The abundant traffic situation is the key of prediction.The ideal database should include traffic situation and typical rule under various influencing factors,but redundant data will lead to increase of computation and increase of search time.The multi section is classified by hierarchical clustering method,which is similar to the data of cross section,and the data of different sections are separated.It ensures the richness of typical data and reduces the time of collecting.(4)Fusion distance study.In the basic algorithm of KNN,Euclidean distance is used as the basis for judging the similarity of samples,but in traffic flow sequence,the way of judging can only measure the proximity and similarity of sequences.The fusion distance,which combines the advantages of both Euclidean distance and cosine distance,can better find out the similar traffic flow sequence in history.(5)Study on optimal parameter value.The window length M and the number of the nearest K are only two preset parameters of the algorithm.It is very important to find the optimal common situation traffic flow to reduce the prediction error.K determines how many traffic days are chosen to predict future traffic,and the M value determines how long the current view measurements are chosen to match.The cross validation method is used to verify the prediction effect of two parameters under different combinations.The curve of the prediction result varying with the value of the parameter is obtained,and the optimal parameters are selected.(6)Prediction experiment.Taking the urban expressway as the prediction object,the search model of common traffic situation and the classical KNN algorithm and the ARIMA algorithm are compared.From the prediction results,the prediction accuracy of the algorithm is 1-4 percentage points better than that of the classical KNN algorithm and the ARIMA model.The MAPE of working days and non working days is about 10%,and the optimum is as low as 9.2%.The experimental results show that the algorithm has high accuracy and strong adaptability.