The bulk insertion of M-tree

• Main Authors: Kuan-Yi Lee, 李冠儀
• Other Authors: Keh-Ning Chang
• Format: Others
• Language: en_US
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/10927301387948016982

碩士 === 國立暨南國際大學 === 資訊工程學系 === 98 === Recently, more and more applications consider the similarity when searching in the database of multimedia data. Because the multimedia does not have a natural order, a distance function is defined by experts to measure the similarity between any two of them. The set of data along with the distance function form a metric space which is so large that a structure should be induced to make insertion and query be done efficiently. The M-tree is such an indexing structure. Since being introduced in 1997, it becomes a paradigm of metric access methods (MAMs) and many extensions of the M-tree storage structure have been developed. The original insertions are done one by one but it can be more efficient if the set of data is inserted as a bulk. Bulk insertion is a bulk operation which inserts a set of data to a non-empty index structure. There are many researches about the bulk insertion on R-trees which process only multi-dimensional objects, but few works on the M-tree structure which is similar to R-tree. We investigate the strategies of the bulk insertion of the M-tree structure. And we conduct experiments to compare different strategies for inserting a set of data to an M-tree.