Multidimensional sequence visualization on a 2D event to sequence correlation matrix

In this study a 2D event-to-sequence floor matrix visualization method and a tool was developed for visualizing multidimensional sequential patterns. In practice, about 200 sequences can be efficiently displayed at once, and sequences can be as long as 20 nodes. Nodes (e.g., instances in time) may consist of up to 30 events (i.e., transactions). Each event may contain up to 20 attributes, which encode more detailed information about each individual event with colors. The amount of sequences and events is limited by the display resolution, while the number of nodes and attributes is limited by the ability of the human eye in utilising colors in an abstract image. The user can drill down from the initial overview to more detailed views and explore data interactively by using sorting, brushing, zooming and selection functions. Two applications, e-commerce transaction and telecommunication traffic, were used to experiment with the tool. The preliminary results indicate that the tool is useful in revealing hidden relationships in multidimensional sequential data.