Mining Census and Geographic Data in Urban Planning Environments

Urban planning is a knowledge-intensive activity. It acquires useful knowledge by analysing jointly socio-economic data and topographic maps. When supported by computers, this preliminary information gathering triggers a knowledge discovery process, which often consists of exploratory data analysis tasks. Advances in georeferencing have caused among the other things a growing demand for more powerful exploratory data analysis techniques. We resort to the field of spatial data mining and propose the task of mining spatial association patterns/rules, i.e. frequent associations between spatial objects, as a means for data exploration. Strong points of the proposed technique for this task are the power of logics as a knowledge representation and reasoning means and the capability of discovering associations at multiple levels of description granularity. This enables new interesting exploratory tasks for spatial data. The technique has been implemented in SPADA and tested on census and geographic data of Stockport, one of the ten metropolitan districts of Greater Manchester, UK. We show results obtained by applying SPADA to a problem of urban accessibility of the Stepping Hill hospital in Stockport.