Header menu link for other important links
Combining spatial and relational databases for knowledge discovery
, A Roberts S
Published in

The advances in the field of data warehousing have led to the creation of huge integrated (usually relational) databases. Organisations that have access to these databases are now using data-mining/knowledge discovery tools to discover hidden relationships between entities of interest to them. At the same time, advances in the field of Geographic Information Systems have led to the availability of equally huge spatial databases, and again data-mining/knowledge discovery tools and techniques have been applied to spatial databases in recent years. There is obviously a synergistic effect in applying knowledge discovery techniques to a combination of these two kind of databases. Example questions are: Is there a correlation between the dietary habits of people and the distances of their home from the sea; Do people change their shopping habits depending on local weather conditions (i.e. - shop closer to home or in covered arcades when it snows); 'Do people who live close to sea-beaches and buy pet food go on fewer holidays'. Such correlations could be discovered in a spatio-temporal database that holds the non-spatial attributes related to sales or (with some difficulty) by querying an appropriate spatial database and a non-spatial database that holds temporal sales data and locational references (e.g. postcodes). Though the first option is the more promising one, there are few - if any - such databases available. The second option poses a problem because of the data being located in two kinds of data structures in two databases. One way to overcome this difficulty would be to copy the required information from one database to the other. However, our aim is to discover hitherto unknown relationships based on the database, so we do not know a priori what data to copy from one database to another. Furthermore we may need to consider several attributes together (as in the case of the third question above) in order to discover interesting correlations.

About the journal
JournalProceedings of the first International Conference on Geo-Computation, Leeds
Open AccessNo