Data Processing Platform Construction And Forecast Realization Of Traffic Flow Based On The WEKA

With the development of detection technology and information acquisition system, huge amounts of traffic information have been acquired, with its related technical booming, especially database technology and data mining technology, and made a lot of theoretical research. In the field of transportation, how to make the most of these techniques into the obtained traffic information effectively has become a hot topic among researchers, aiming to analyze traffic flow efficiently for the engineer. Based on these requirements, a data processing platform is established in this thesis covering the process from the traffic flow data storage to the data processing analysis, which means the SQL Server is connected with WEKA successfully through the software interface. The process of the traffic flow analysis is studied on this platform. K-means cluster is combined with BP neural network, realizing the traffic flow forecasting. The whole research work is mainly from the following aspects.Firstly, Comparing the current popular three data mining tools, WEKA with advantages of built-in database interface and machine learning algorithm is worked as the main data mining tool in this research. The interface connecting the WEKA and SQL Server 2005 is developed, achieving the connection between the database and data mining tool. The one-stop platform is established from traffic flow data storage to data processing analysis.Secondly, the process of traffic flow data processing based on the platform is designed, which means the whole process is achieved ranging from data storage, data preprocessing to traffic flow forecasting. The database algorithm, data compensation algorithm, filtering algorithm are discussed. The traffic flow forecasting algorithm combining the K-means cluster and BP neural network is proposed.Finally, the traffic forecasting experiment is carried out according to the traffic flow data in a big city. It is proved that this platform enables to implement traffic pretreatment and forecast with better performance. Relying on this platform, the promptness and convenience of traffic flow data process is verified.