Research And Application On Spatial Clustering Methods Considering Rule-bounds

Spatial clustering is one of the most important methods in the spatial data mining field. Butin recent years, only very little research is done in spatial clustering, which directly leads to thedecline of its proportion in practical applications. Different from classification with rules,Clustering analysis has an irreplaceable advantage, which is that it can find significant spatialclustering structure directly from the spatial database and does not need background knowledge.On the basis of the scholars’ previous work, this paper studied the spatial clustering withrule-bound. That is, this paper addicted spatial rule-bound, non-spatial attributes rule-bound, andorientation rule-bound, to the conventional spatial clustering in order to make the spatialclustering suit people needs better.This paper contains the following contents:Firstly，this paper analyzied the background, significance and relative theoretical andtechnical basis of the spatial data mining and spatial clustering theory. Started from the analysisof the background and significance of spatial clustering analysis, this paper studied the researchstatus both home and abroad in order to get the main problems existed in the nowadays study.Secondly, this paper studied the basic laws in spatial clustering analysis and spatial rankinganalysis. Then it gave the basic framework and main algorithm of spatial clustering analysis.Also combaining spatial clustering analysis, this paper gave the basic process of spatial rankinganalysis and its paractical significance in combining these two analyses. Besides, this paperanalyzed the need that we put forwad the rule-bounds. Futherly this paper gives the relativequestions of the rule-bounds.Thirdly, besids the two normal types of the rule-bounds, spatial rule-bound and noe-spatialattribute rule-bound, and this paper raised another rule-bound, the orientation rule bound. Thesethree rule-bounds are different in metric, and also different in the way attaches to the spatialclustering. This paper realized the spatial clustering in the grid space. And through doingmathmaticl calculating and conversion for the three rule-bounds, this paper can attach all theserule-bounds to the spatial clustering at the same time, thus realized the co-influenced spatialclustering of three rule-bounds.Forthly, like the spatial clustering, ranking analysis can also dig out a potential law hidden inthe data. And the difference is ranking analysis needs to set ranking indicators and their weights.Thus this paper studied how to do spatial ranking analysis for the spatial clustering results withthe help of the rule-bounds. On the basis of traditional non-spatial attribute spatial ranking, this paper proposed a new spatial ranking method, the oriention spatial ranking methods, as well asits realization method, which had made the non-spatial attribute rule-bound and orientationrule-bound participated the spatial ranking analysis simultaneously. What’s more, did unifiedmathematical transformations for the two spatial ranking types, thus this paper can implement aspatial ranking method under the influence of both two ranking factors.5. This paper designed an experimental prototype system to realize the spatial clustering withrule-bounds, the ranking analysis based on the spatial clustering result in order to verify thetheory and methods proposed in this paper. Beyond this, the paper found a new applicationdirection for the spatial clustering. That is when combaining spatial clustering and spatialranking together, we can apply them to the interest points selection in the electronic map. Andconsulting the obstacle distance calculating method, this paper realized the path planning wayunder the real-time traffic environment.