基于矩阵伪投影策略的频繁项集挖掘方法

Mining Frequent Itemsets Based on pseudo-Projection of Array

挖掘频繁项集是数据挖掘应用中关键的问题。经典的FP-growth算法利用FP-tree有效的压缩了数据集的规模,但是在挖掘过程中需要反复递归构造条件FP-tree成为限制算法效率的瓶颈。本文通过将FP-tree映射成矩阵,通过在矩阵自身上进行伪投影得到条件模式阵,避免了递归构造FP-tree,从而节约了内存消耗和计算时间。

It is key point of data mining application mining frequent itemsets. Classic frequent itemsets mining algorithm FP-growth compresses the scale of dataset effectively using FP-tree structure, But it has own bottleneck that for getting complete fre- quent itemsets it need build conditional FP-tree recursively in the mining process. This paper proposes a new frequent itemsets mining algorithm that maps FP-tree structure into FP-array and mines upon it. In the mining process, this algorithm can avoid building conditional FP-tree. So, it saves time and memory very much .