| Spatial data mining and knowledge discovery is the process of spatial data from the concentrated extract implicit spatial information. Spatial association rule is an important conposition of spatial data mining and knowledge discovery. The operation process of spatial association rule discovery includes the calculation and analysis of many spatial data. In order to better find spatial association rules, spatial association rule mining is more reasonable. This paper presents expression of geographic space and background knowledge for constraint set and the constraint set guidance of spatial association rule mining method.With the development of spatial mining technology and the accumulation of a large number of spatial data sets, the mining technology of spatial association rules has become more and more attentioned. With a set of spatial data sets, spatial association rules mining can find a certain kind of high frequency complex spatial relation among geographical entities in the geographical space. For example, a plant always exists in a geographical environment with some kind of mineral products. Spatial association rules are different from other common mining methods, and the spatial association rules are dependent on the specific spatial and spatial relationships of entities in space. Therefore, the effectiveness of spatial association rules mining is greatly dependent on the processing of spatial attribute and spatial relation. For example, spatial objects have various spatial relations, such as topology, orientation and distance, in which the shape is a point, line, plane and so on. Therefore, the spatial data have been added to the basic properties of ordinary data, and also with a variety of spatial properties (relationship). Spatial data are similar and different because of the characteristics of the complex space environment The distance constraints between properties of spatial objects and their relations are nonlinear. A particular area with its constraints is different with other regional conditions and the complexity of spatial data causes the complexity of spatial association rules.Till now a lot of spatial association rules algorithms have been proposed, but the results obtained are not very satisfactory. One of the reasons for it is few space constraints. These algorithms are lack of analysis of regional background knowledge constraints and supervision of spatial association rules mining process.The main problem of the spatial association rules mining includes following few aspectses:1. Spatial data is complex.2. The background knowledge in mining process had not given enough attention. In the process of mining spatial predicates by spatial analysis, it is not enough to focus on the algorithm.In some scenes, spatial association rules discovered the flaw and the wrong.Based on the above problems and reasons, this paper studied ontology N3 expression of spatial association rules in the process of spatial data and spatial association rules mining algorithms.The expression based on the establishment of various constraint rules set merged with inference of these rules injected into spatial association rules mining process to improve the quality of the results set of spatial association rules. The concrete research contents areas follows:(1) Based on the basic principle of spatial association rules, it was introduced that the problem of association rules algorithm. First of all, according to the characteristics of spatial data and by analyzing the vector data about point,line and polygon, a unified approach based on polygon is proposed to deal with the three basic formats classification features which is called Multi-type Object Spatial Apriori(MOSApriori).Then, the overlay analysis between the polygons is set to get the vertical relationship of spatial predicates. The points of intersection between the polygons have been calculated by the algorithm. If there is without points of intersection between the polygons, the analyse of the horizontal spatial correlation predicate were conducted on the raw data. According to calculation method of the support and confidence in spatial space, weakly spatial association rules are filtered. Finally, it is analysed that the shortcomings of MSOApriori algorithm. According to illustrate the problems of background knowledge constraint in algorithm, it is pointed out that spatial association rules may not be found because lack of the constraints.(2) Based on the basic theory of ontobgy and the ontology technobgy, knowledge expression of spatial association rules are discoved base on the data and algorithm MSOApriori. MOSAprioriO ontobgy is designed in the application framework.An ontobgy application framework MOSAprioriO includes two kinds of concept hierarchy tree SpatialThingFeature and MOSAprioriAlgorithm. Two kinds of ontology object relationship DataRelation in SpatialThingFeature and AlgorithmRelations in MOSAprioriAlgorithm are designed properties, giving several examples to illustrate in-depth analysis. The main design idea of MOSAprioriO is how to expresses knowledge to realize mining spatial association rules with the method of constraints.(3) In the framework of MOSAprioriO ontobgy application, it has been introduced that the spatial association rules extraction with constraint algorithm C-MOSAprioriO(constraints Multi-type object Spatial the Apriori ontology algorithm) based onMOSAprioriO.It gives the MOSApriori ontology reasoning process, methods and examples. This paper according to the MOSAprioriO establish the constraint rules for the use of the spatial association rules.MOSAprioriO and constraint rules have been set to consisting of geographic information background knowledge to guide the spatial association rules extraction.This paper is propsed a new algorithm based on MOSAprioriO to represent knowledge.The paper detailed elaborated the work, the principle of the method, and a module algorithm for data clustering analysis with specific examples of constraint analysis.(4) Based on thought of C-MOSAprioriO, demo for C-MOSAprioriO has been developed.In order to carry out the experiment and evaluation, a collection of data and the choice of evaluation index have been actually tested. C-MOSAprioriO and MOSApriori have been compared.Because the support of background knowledge of geography, the effect of C-MOSAprioriO is better than that of MOSApriori. Interest rate of association rules in MOSApriori is much lower than that in C-MOSAprioriO. |
PDF Full Text Request