Development of a Statewide Transportation Data Warehousing and Mining System under the Louisiana Transportation Information System (LATIS) Program [Final Report]

2008PDFTech Report42804-1SSData miningInformation systemsTraffic dataData warehousesIntelligent transportation systemsSoftwareInfrastructureTraffic data warehouseintelligent transportation systemsdata miningdatabasesarchived data management systemsLouisianaTulane University. Department of Electrical Engineering and Computer ScienceBuckles, Bill P.Ishak, SherifSmith, StephanieTulane University. Dept. of Electrical Engineering and Computer ScienceLouisiana Transportation Research CenterUS Transportation CollectionMore jurisdictions including states and metropolitan areas are establishing traffic management centers to assist in reducing congestion. To a lesser extent, these centers are helpful in providing information that assists engineers in making such adjustments as signal synchronization or road improvements. However, the main traffic management center function is real-time decision making for freeways and surface streets.Planning and modification of a traffic network is best pursued in a context that includes large historical data. Understanding traffic \u201cbehavioral\u201d properties is the domain of numerous technologies such as data mining which depends on the presence of large amounts of historical data. To these ends, the Louisiana Transportation Research Center (LTRC) commissioned this study for the design of a data warehousing/data mining system that, while limited to Baton Rouge, will serve as a statewide model. Few traffic-oriented data warehouses exist in the U.S. The methodology employed in designing the Baton Rouge, Louisiana warehouse included visiting many of them and collecting sample data from Baton Rouge sensors. Advisory and stakeholder committees were formed to give advice on the base applications. Base applications are the ones recommended for the initial inclusion in the warehouse. The applications were traced back to the data, resulting in some of them being dropped or modified to suit the data and its quality that was available. From that juncture, the data was tracked forward again to the applications, modeling the transformations necessary. This transformation set constituted the design. Chosen applications, in addition to data mining, included several variations of performance measuring and hydrowatch. The latter is unique among traffic warehouses and is particularly appropriate for the region and State.The data warehouse design consists of a system with three stages \u2013 extraction/transformation/loading (ETL) of source data, main storage of the warehouse, and client workstation software. ETL consists of acquiring, cleansing, formatting, merging, and purging of the source data. Much of this stage entails data quality checking and the report addresses this aspect. Main storage is organized around a star schema also called a multidimensional data cube. This separates the static data such as sensor location from dynamic data such as lane occupancy. This design assumes a one-way data flow, input from the sensors and output to client workstations or other media. This approach is codified commercially as online analytical processing (OLAP). Many warehouses stop at the main storage phase (the \u201cdump and run\u201d model). Here, the solution to key client phase issues, interfaces to GIS and linear referencing are given.Infrastructure and a marketing plan are given. The key infrastructure decision is determining the warehouse\u2019s physical location \u2013 at LTRC, at a university, or at a private/public concern. All three variations were discovered in site surveys. Marketing consists of addressing various segments beginning with those who are inclined to add value to the system. These are the engineers and planners but at some stage university researchers and the general public must be given access and be convinced of the value that can be obtained.93