摘要： 针对现有频繁邻近类别集挖掘算法存在重复计算和冗余邻近类别集的问题，提出一种依赖于真子集的频繁邻近类别集挖掘算法，适合在海量数据中挖掘空间对象的频繁邻近类别集。该算法用析构法建立邻近类别集数据库，用产生邻近类别集真子集的方法计算支持数，实现一次扫描数据库提取频繁邻近类别集。算法无需产生候选频繁邻近类别集，且计算支持数时无需重复扫描，从而达到提高挖掘效率的目的。实验结果表明，在海量空间数据中挖掘频繁邻近类别集时，该算法比现有算法更快速有效。

关键词: 邻近类别集, 真子集, 析构, 递增搜索, 空间数据挖掘

Abstract: Aiming at the problems that the presented frequent neighboring class sets mining algorithms have repeated computing and superfluous neighboring class sets, this paper proposes an algorithm of frequent neighboring class sets mining dependent on proper subset, which is suitable for mining frequent neighboring class sets of spatial objects in large data. The algorithm uses the way of destructor to create database of neighboring class sets, and uses the way of generating proper subset of neighboring class sets to compute support, it only need once scan database to extract frequent neighboring class sets. The algorithm improves mining efficiency by two approaches. One is that it needn’t generate candidate frequent neighboring class sets, the other is that it needn’t repeated scan database when computing support. The result of experiment indicates that the algorithm is faster and more efficient than presented algorithms when mining frequent neighboring class sets in large spatial data.

Key words: neighboring class sets, proper subset destructor, ascending search, spatial data mining

中图分类号: 

• TP311.52