Abstract: As one of the important research in spatial data mining, the spatial co-location pattern mining attracts more and more attention. In practical applications, in addition to the spatial information, the spatial features usually contain the attribute information which is important for the decision-making. However, the previous works stressed the spatial information only. Firstly, based on the fuzzification of attribute information, this paper defines the related concepts, including fuzzy feature, fuzzy co-location pattern, etc. Secondly, similar to the related concepts of traditional co-location mining, this paper defines some concepts of fuzzy co-location pattern, such as table instance and participation index, etc. Then, this paper proves the downward closure property of fuzzy co-location pattern, proposes a basic mining algorithm, and puts forward two pruning strategies so as to improve the mining performance. Finally, by extensive experiments, the efficiency and effectiveness of the algorithms are verified.

Key words: spatial co-location pattern, fuzzy attribute, fuzzy co-location pattern, pruning

摘要： 空间Co-Location模式挖掘是空间数据挖掘的一个重要研究方向，正受到越来越多的关注。在实际应用中，空间特征不仅包含空间信息，还经常伴随着属性信息，这些属性信息对决策和知识发现有重要意义。然而现有的Co-Location挖掘方法只强调特征的空间信息，忽略了其属性信息。基于对属性信息的模糊化处理，定义了模糊特征和模糊Co-Location模式等概念。类似于传统空间Co-Location模式挖掘中的相关概念，定义了模糊Co-Location模式的表实例和参与度等概念。在证明模糊Co-Location模式的向下闭合性质的基础上，设计了一个基本挖掘算法。为提高算法的可伸缩性，提出了两个剪枝方法。在合成的和真实的数据集上进行了大量实验，验证了基本算法及其改进算法的效果和效率。

关键词: 空间Co-Location模式, 模糊属性, 模糊Co-Location模式, 剪枝