Mining high utility itemsets from multiple databases

Abstract

In the past, many algorithms have been developed to efficiently mine the high-utility itemsets from a single data source, which is not a realistic scenario since the data may be distributed into varied branches, and the discovered information should be integrated together for making the effective decision. In this paper, we focus on developing an efficient algorithm for synthesizing the mined high-utility itemsets from different sources. A baseline algorithm is first designed and two criteria are then developed to verify whether the designed algorithm is efficient to generate the same number of the high-utility itemsets as the batch-processed algorithm. Experiments are then shown that the designed algorithm has good performance for rule synthesization.