Abstract: Aiming to these shortcomings from the present frequent closed itemsets mining algorithms, this paper proposes an algorithm of frequent closed itemsets mining based on granular computing. The algorithm uses the varying mixed radix number to generate candidate itemsets, and avoids adopting the complex data structure to reduce the memory and the CPU overhead. And it uses divide and rule for granular computing to compute the support of frequent closed itemsets, and avoids reading repeatedly the database to reduce the computation complexity and I/O overhead. These experimental results indicate that the algorithm is faster and more efficient than these classical mining algorithms for frequent closed itemsets.

Key words: frequent closed itemsets, granular computing, data mining

摘要： 针对现有频繁闭项目集挖掘算法存在的不足，提出了一种基于粒度计算的频繁闭项目集挖掘算法。通过混合进制数的变化来生成候选项目集，避免使用了复杂的数据结构，减少了内存和CPU的开销;利用粒度计算的分而治之思想来计算频繁闭项目集的支持度，避免了多次重复扫描数据库，减少了计算复杂度和I/O开销。实验结果表明该算法比经典的频繁闭项目集挖掘算法快速而有效。

关键词: 频繁闭项目集, 粒度计算, 数据挖掘