Visual exploration of large traffic database using traffic cubes

Advances in data collection tools and computation abilities result in a significant increase of capabilities for both collecting and generating data. The consequent growth of available data generates an urgent need for techniques to transform the massive amount of collected data into useful information. This thesis develops traffic-cube based visualization techniques, and demonstrates the efficiency of those techniques as an approach of mining traffic flow data. The basis of these techniques is a data mining technique known as data cube, which is an array of values defined across multiple measurement dimensions that allows rapid, user-interactive aggregation and cross-tabulations of data. The traffic cube is an extension of the data cube that facilitates these operations and is usually used to organize and manage traffic flow data. In this thesis, the traffic cube is defined by three dimensions: the clock time (T), the calendar date (D), and the spatial location (S) of the recorded traffic count. For each dimension, a hierarchy is designed for data aggregation; and we visualize the aggregated two-dimensional slices of the original traffic data cube to extract useful information about traffic situation.