Font Size: a A A

Research On Spatial Co - Location Pattern Mining Based On Parallel Computing

Posted on:2015-03-10Degree:MasterType:Thesis
Country:ChinaCandidate:F Z HeFull Text:PDF
GTID:2208330431969161Subject:Computer technology
Abstract/Summary:PDF Full Text Request
Recently,various kinds of spatial data has increased rapidly. Facing these mass spatial data, spatial data mining technology has become an effective method to analyze and make use of spatial knowledge. The personal high performance computers and multi-core system computers are widely used in the field of data mining, but current software technology can hardly make full use of these high performance hardware. The parallel computing has become an effective method to solve the contradiction of non-coordinated development between hardware and software. Spatial co-location pattern mining researches a set of spatial features whose instances are frequently located in space. Co-location pattern mining is an important research direction of spatial data mining, and many related algorithms have been proposed in this field. However, these algorithms are all based on serial computing, so they are time-consuming, especially when deal with mass data. As we know, parallel algorithms can improve the efficiency of solving the same problem through working on multiple computers simultaneously. This article studies spatial co-location pattern mining on the basis of parallel computing, and we propose a novel co-location pattern mining algorithm based on parallel computing. We did lots of experiments on three data sets. Experiments showed that the novel algorithm based on parallel computing can greatly improve the efficiency of spatial data mining and provide a new method to mine mass spatial data in a fast and effective way.First, this part introduces the concept of parallel computing technology and discusses the problems and shortcomings of parallel computing.Secondly, we introduce the research and related concepts in spatial co-location pattern mining. We discusses three method of mining co-location patterns, included based on the minimum participation rate, mining maximal participation rate and mining complex patterns. This paper elaborates the process of join-based algorithm.Third, this part discusses the related problems in spatial co-location pattern mining based on parallel computing. We put forward the data partition algorithm based on middle axis. Based on the above steps, we propose a novel spatial co-location pattern mining algorithm based on parallel computing.Fourth, we do a lot of experiments on three data sets and study how the distance threshold, the participation threshold and the number of instances influence the efficiency of the algorithm. We discusses the relationship between numbers of PC and the efficiency of the algorithm. We also summarizes the related rules. The correctness and scalability of the algorithm is proved in this paper. We analysis the time complexity of join-based and PC-join-based algorithm.At the last, we summarize the main contents of this thesis and give the outlook of the future research.
Keywords/Search Tags:Spatial co-location pattern mining, Data partition, Parallel computing, Parallelalgorithm
PDF Full Text Request
Related items