In recent years,the rapid development of intelligent transportation system has become a reliable guarantee to people's daily outing.Real-time prediction of short-term traffic flow is an important basis for traffic control and vehicle-induced,and is also one of the key research in intelligent transportation system field.The development of big data leads to the explosive growth of traffic data,and how to achieve a real-time and accurate prediction result becomes a new problem under the background of huge amounts of data.In this paper,deep analysis,research and experiment are carried out for real-time prediction of short-term traffic flow.Related research works are as follows:(1)In order to improve the accuracy of mining and prediction,the collected traffic flow data should be pre-processed.In this paper,related pre-treatment includes missing filling,error correction,symbol discretization and other ETL processing.(2)Under the background of big data,this paper proposes a real-time traffic flow data mining algorithm of frequent closed patterns—TP-Moment.The Topology parallel model is applied to improve the traditional Moment algorithm.In the experiment of large data set,the algorithm shows good mining accuracy and improves a lot in time and space performance.In short,it can greatly satisfy the prediction accuracy and real-time requirements.(3)On the basis of TP-Moment,the prediction model based on historical frequent patterns is proposed.By mining the frequent patterns of historical traffic flow and combine with real-time traffic information based on neighborhood matching principle,the future traffic flow condition can be forecasted.Experiment dataset shows the high prediction accuracy of model,and the prediction model is effective and feasible. |