Research Of Multidimensional-Index On Intelligence Metadata

With the rapid development of intelligence technology in China, the intelligence collected by intelligence department through many different ways is increasing everyday. The intelligence is continuously delivered to the related department to be analyzed and used by people. But the intelligence is complicated that they are connected to each other, so the intelligence should be integrated analyzed and deeply mined when we use them. This may increase the difficult of the intelligence management.The function of the intelligence metadata is to achieve the efficient management and rapid query of the intelligence. Considering the data structure of the intelligence metadata, we can realize the efficient management of the intelligence by establishing a global, common and complete management model of the metadata. So people can rapidly, conveniently and exactly query the intelligence.Under the background of the intelligence management, this thesis has deeply analyzed the multidimensional-index technology and related query algorithm by applying the management of the intelligence metadata in the multi-space, this thesis has finished the following works:(1)We provide a deep analysis of the intelligence metadata. After a brief survey of several criterions, we first present the characters of intelligence metadata, then we apply a kind of binary dimensional hierarchical encoding method to reflect the hierarchical property of the metadata and present a method of mapping the intelligence metadata into the multi-space.(2)With the considering of the characters of the intelligence metadata this thesis proposes a new high-dimensional index: VAP-tree. It is a multidimensional-index of a combination of the VA-file and pyramid technology, we first use a compact, geometric approximation method which comes from VA-file to process each vector and then use the pyramid technology to construct a B+ tree index for all of the approximation values. Then, we discuss the algorithm of point query, range query and kNN query on our tree. Several theoretical and experimental studies manifest that VAP-tree over-performs the VA-file and pyramid technology in high-dimensional situation.(3)Finally, we discuss a prototype of the integration management system of intelligence metadata, we present the practical meaning and the applying method of using multi-index in the management of the intelligence by giving a analysis of every function module.