Font Size: a A A

Compact Co-location Pattern Mining Fuzzy Object

Posted on:2015-01-15Degree:MasterType:Thesis
Country:ChinaCandidate:F S WenFull Text:PDF
GTID:2260330431967292Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
Spatial co-location pattern represents a subset of spatial objects, and their instances frequently associated in space. Currently it has been lots of research for spatial co-location pattern mining, mainly including two directions:spatial co-location pattern mining on determination data and uncertain data, while the researches of spatial co-location pattern mining on fuzzy objects are rare. The fuzzy object data exists everywhere in our daily life and these mining algorithms of determination data and uncertain data do not apply to the data of fuzzy objects, so we need work out a new algorithm to carry on the spatial co-location pattern mining of fuzzy objects. This article mainly studies the compact co-location pattern mining algorithm of fuzzy objects and its improved algorithms. The content of this article is shown as follows:Firstly, this article outlines co-location pattern mining and the achievements of spatial co-location pattern mining of fuzzy objects data, and describes related concepts, properties and theorems of spatial co-location pattern mining, and related concepts and theorems of fuzzy theory.Secondly, this article put forward the Mevent-tree algorithm to mining maximum co-location patterns for fuzzy objects. The candidate patterns gotten from the spatial object trees constructed for each objects, and then we build the HUT trees for these candidate patterns. we depth-first search maximum co-location patterns from the maximum-size candidate patterns to the size-2candidate patterns in the HUT tree, and pruning HUT tree after getting maximum co-location patterns. Then two improved algorithms were put forwarded. Extensive experiments show the performance of Mevent-tree algorithm and its improved algorithms.Then, the U-HUT algorithm was proposed to mining top-k closed co-location patterns for fuzzy objects. The U-HUT trees were builded after obtaining the candidate patterns, and breadth-first search the U-HUT trees from the size-2candidate patterns to maximum-size candidate patterns for top-k closed co-location patterns. After that, extensive experiments show the performance of U-HUT algorithm and its improved algorithm.Finally, this article gives a brief summary of all the work, and points out its disadvantages and future research direction.
Keywords/Search Tags:fuzzy objects, maximum co-location patterns, top-k closed co-locationpatterns, fuzzy participation rate
PDF Full Text Request
Related items