Efficiently Mining Maximal Frequent Patterns from Traversals on Weighted Directed Graph Using Statistical Theory

To solve the problem of mining weighted patterns with noisy weight from traversals on weighted directed graph (WDG), an effective algorithm, called SMaxWFPMiner (Statistical theory-based Maximal Weighted Frequent Patterns Miner), is proposed. The algorithm undergoes two phases to discover MaxWFP from the traversals on WDG. In the first phase, it adopts the weight's confidence level (CL) to remove the vertices with noisy weights, which reduce remarkably the size of traversal database (TDB). In the second phase, incorporating the maximal property with weight constrains, it exploits a weighted FP-tree approach to reduce effectively search space and extract succinct and lossless patterns from weighted graph TDB. Experimental comparison results show that the algorithm is efficient and scalable for mining MaxWFPs based on traversals on the WDG.