Spatial Data Mining

The targets of our project is to explore efficient algorithms for finding frequent patterns in spatial context from data in relational databases and / or web pages that contain spatial information such as addresses. Following are examples of spatial data mining functions that we have developed so far.

• Neighboring Class Sets

This function finds frequent neighboring class sets. Objects of each instance of a neighboring class set lie in close to each other. For example, from access logs of a location based service, it may find a frequent neighboring class set, "timetable" service and "ticket" service are frequently requested in close to each other, as circled in the map.

• Optimized Distance / Orientation

This function computes distance and / or orientation range from point objects of a class. The distance and / or orientation range optimizes a user specified criterion. For example, it may find "X meter distance from an ATM" maximizes crime density.

• Optimized Region

This function computes pixel grid region that optimizes a user specified criterion. For example, it can compute a region that maximizes crime density.

(You can find our previous works on Data Mining in this page.)

Research items

• Spatial OLAP algorithms for aggregating database based on geographical classifications such as for each town.
• Spatial data mining algorithms and facilities for extracting association rules on geographical context from databases that contain spatial attributes such as addresses.