Research On Ontology Assisted Geospatial Association Rule Mining

Spatial association rule mining, which is an important part of spatial data mining andknowledge discovery, is mainly interest in mining connotative rules like spatial object structureand connections between spatial or non-spatial properties from GIS database. In existential studyabout spatial association rule mining, many researchers pay more attention to data andalgorithms but little to apriori knowledge holed by user and mining task themselves, thus someacademicians proposed rule mining methods with constrains and apriori knowledge. As it’s thefoundation of semantic analysis of semantic web, describing logic based ontology constructerhas distinctness advantage in user knowledge’s expression and therefore be introduced into datamining to represent mining task relevant knowledge. Most ontology assisted methods focus onbuilding data mining ontology and controlling mining workflow, but ignoring optimizing themining content. After analyzed methods of spatial association rule mining, their superiority anddeficiency, this paper proposed a new ameliorative way to syncretize ontological representingknowledge and mining process.The main research contents of this paper are as follows:Firstly， this paper analyses research the nowadays status of the spatial association rulemining, and expatiates the relevant theory as well as the building principles of ontology at thesame time. Then it introduces four dimensions of―object of discourse‖, which build the bedrockfor this paper to discusses the conversion from―object of discourse‖to ontology and proposes afive steps way in semantic contracting in order to build an ontology.Secondly, this paper analyses the existent problem in pretreatment phase of spatial datamining, and more attention is paid to data cleaning and data reduction. Considering that data isincomplete, this paper deals the data in a new way that it through calculating semantic similarityamong the ontology conceptions which uniquely related with attributes in database to obtainameliorative dataset, and thus the pretreatment data space is reduced. By using conceptualsystem expressed in ontology to choose appropriate extent or arrangement, this paper proposesan effective way to deal with data reduction.Thirdly, this paper expounds geographic dependences phenomenon existed in spatialassociation rule mining and proves that closed frequent sets mining cannot eliminate thesedependences. Then, after introduces some relevant conceptions of Lattice and its generator, itdescribes relations between generator and geographic dependences and expatiates how to usinggenerator to eliminate geographic dependences. In order to realize the application of ontologysemantics, this paper proposes an algorithm called ontology semantic based optimize frequentgeographic pattern mining (OS-OFGP), and its validity is proved through the example of Changsha.Last but not least, this paper discusses knowledge representation by ontology as well as themapping between ontology and dataset. Besides, it designs rule schema to obtain interestingrules along with filter operator. Upon defining similarity between rules, it proposes a way thatusing conception similarity of ontology to calculate similarity between rules so as to realize theextraction of the rules. Further more, take Henan as an example, this paper proves the efficiencyof the rules extraction methods.