空间亚频繁co-location模式的主导特征挖掘

Dominant feature mining of spatial sub-prevalent co-location patterns

空间co-location模式是一组空间特征的子集,它们的实例在邻域内频繁并置出现。通常,空间co-location模式挖掘方法假设空间实例相互独立,并采用空间实例参与到模式实例的频繁性(参与率)来度量空间特征在模式中的重要性,采用空间特征的最小参与率(参与度)来度量模式的有趣程度,忽略了空间特征间的某些重要关系。因此为了揭示空间特征间的主导关系而提出主导特征co-location模式。现有主导特征模式挖掘方法是基于传统频繁模式及其团实例模型进行挖掘,然而,团实例模型可能会忽略非团的空间特征间的主导关系。因此,基于星型实例模型,研究空间亚频繁co-location模式的主导特征挖掘,以更好地揭示空间特征间的主导关系,挖掘更有价值的主导特征模式。首先,定义了两个度量特征主导性的指标;其次,设计了有效的主导特征co-location模式挖掘算法;最后,在合成数据集和真实数据集上通过大量实验验证了所提算法的有效性以及主导特征模式的实用性。

The spatial co-location pattern is a subset of spatial features whose instances frequently appear together in the neighborhoods.Co-location pattern mining methods usually assume that spatial instances are independent to each other,adopt a participation rate,which is the frequency of spatial instances participating in pattern instances,to measure the importance of spatial features in the co-location pattern,and adopt a participation index,which is the minimal participation rate of spatial features,to measure the interest of patterns.These methods neglect some important relationships between spatial features.Therefore,the co-location pattern with dominant feature was proposed to reveal the dominant relationship between spatial features.The existing method for mining co-location pattern with dominant feature is based on the traditional co-location pattern mining and its clique instance model.However,the clique instance model may neglect the non-clique dominant relationship between spatial features.Motivated by the above,the dominant feature mining of spatial sub-prevalent co-location patterns was studied based on the star instance model to better reveal the dominant relationship between spatial features and mine more valuable co-location patterns with dominant feature.Firstly,two metrics to measure feature’s dominance were defined.Secondly,an effective algorithm for mining co-location pattern with dominant feature was designed.Finally,the experimental results on both synthetic and real datasets show that the proposed mining algorithm is efficient and the co-location pattern with dominant feature is pratical.