Frequent Itemsets Mining with VIL - Tree Algorithm

The aim of this paper is to develop a new mining algorithm to mine all frequent itemsets from a transaction database called the vertical index list (VIL) tree algorithm. The main advantages of the previous algorithms, which are frequent pattern (FP) growth and inverted index structure (IIS) mine, are still useful in a new approach as database scanning only done once, and all frequent itemsets are mined without generating candidate itemsets and the changing in the minimal support threshold is not affected the data structure. IIS Mine was proposed to reduce the recursive mining steps, nodes construction, and the size of the trees. However, IIS mine has some drawbacks when the small transaction sets are contributed to early trees, so sub trees are of the result. To overcome this problem, VIL Tree has been proposed to mine large transaction sets to get the early long size of frequent itemsets. This is useful when many subsets of frequent itemsets are found, and from it recursive mining steps, nodes, and sub trees are reduced. The performance of VIL Tree has been tested with reference to FP growth and IIS Mine. The experimental results demonstrate that VIL Tree provides better performance than the two comparative algorithms in terms of run time and space consumption.