Research On R-tree Index Method Based On Chameleon Algorithm Clustering

With rapid development of GIS technology,the spatial database whose main task is to achieve effective storage of spatial data and target is to obtain effecive access,has played crucial roles in various fields.As spatial data bears characteristics of mass,complexity of internal structure and diversity of attribute,therefore,it's a hard issue for the current spatial database field to achieve effective storage of data.R tree index structure is capable of implementing effective storage of high dimensional mass spatial data and as there's still proximity in physical space between data,it is mainly applied in commercial database.In view of shortcomings of the current R tree index structure,this paper has conducted research in accordance with the following three parts:First of all,aiming at the deficiency of the traditional R tree construction and splitting method,in this paper,it has been preprocessed in combination with the clustering algorithm,so that a technique which makes mass generation of index structure available has come into being.By using characteristics of high similarity within clusters and low similarity between clusters of clustering results,the MBR area of node can be reduced,so that overlapping is avoided and efficiency of algorithm is improved.The similarity between nodes that through clustering is rather low,the method of avoiding multi-path search during query process has improved the query efficiency.Secondly,according to high time complexity of Chameleon algorithm,the centroid of cluster is obtained by the method of artificial bee colony,and it is used as the initial solution of the next clustering.The way of using K-means algorithm to conduct next clustering has avoided the influence of arbitrary initial value and noise points on the R tree nodes.At the same time,the time of constructing index structure has been reduced.Therefore,the static R tree bearsproperty of flexibility when dealing with huge data.Finally,according to the complexity of uncertain data storage,in combination with clustering algorithm and the Hilbert curve dimension reduction method,by taking advantages of pruning strategy of the maximum and minimum rectangle,the amount of calculation of integral in algorithm has been reduced and the efficiency of construction Hilbert-R tree has been improved in this paper.The method of using Chameleon clustering algorithm will make the data more compact and data except for root node of full capacity,in this way,space utilization ratio of node is improved.While taking the interconnection between nodes into account,the potential contact between adjacent data can be discovered.