基于粗糙集的局部同位模式挖掘算法北大核心CSCD

通过空间同位模式挖掘可发现频繁发生在邻近位置的事件集合,为揭示地理现象间的共生规律提供重要的决策支持。由于空间同位模式存在空间异质性问题,已有方法不能很好地探测出空间同位模式分布的相近性区域。为此,本文从地理属性的相近性方向探测同位模式的分布区域,提出了基于粗糙集的局部空间同位模式挖掘方法。首先,从全局视角提取不频繁的空间同位模式作为候选的局部空间同位模式;然后,对候选同位模式的实例位置进行处理,将其分布的热点区域属性作为粗糙数据集,借助粗糙集探测局部空间同位模式自然的分布区域;最后,度量在这些局部区域的频繁程度,生成所有频繁的局部空间同位模式。通过试验与应用发现,该方法不仅可以探测局部空间同位模式分布的相近性区域,还能反映同位模式分布区域的地理属性信息。

Algorithm of regional co-location patterns based on rough set

Through co-location pattern mining, we can find the set of events frequently occurring in nearby locations, which provides important decision support for revealing the symbiosis law between geographical phenomena.Due to the spatial heterogeneity of co-location patterns, the existing methods can not detect the distribution of co-location patterns.Therefore, in this paper, we detect the distribution regions of isotopic patterns from the direction of proximity of geographical attributes, and propose a local co-location pattern mining method based on rough sets.Firstly, infrequent co-location patterns are extracted from the global perspective as candidate local co-location patterns. Then, the locations of the candidate co-location patterns are processed, and the attributes of hot spots are used as rough data sets to detect the natural distribution regions of local co-location patterns. Finally, the frequency of these local regions is measured and all frequent local co-location patterns are generated. Through experiments and applications, it is found that this method can not only detect the similarity region of local spatial distribution of the same location pattern, but also reflect the geographical attribute information of the same location pattern distribution region.