Discovering probabilistic weighted frequent itemsets over uncertain data

The uncertain data management and mining is a growing research topic in recent years. To mine more meaningful patterns, some algorithms have considered the importance of every items as a constraint. None of them have been, however, designed to discover patterns in reasonable such as Possible World Semantics (PWS) which has usually adopted. In this paper, we defined the weighted probabilistic of frequent itemsets, which provides a better view on how to obtain the more interesting patterns under PWS. In terms of the concept, a deepth-first algorithm PWFIM is proposed to generate the results, and we also designed a Dynamic Programming method and several pruning methods to further improve the mining performance. We have carried out substantive experiments on real life and synthetic data sets. The results show that the proposed algorithm can be more meaningful and interesting than other data algorithms. We also evaluated the performance of the algorithm at runtime, consumption of memory, and number of patterns.