Analysis of Methods for High Utility Pattern Mining without any Threshold Setting

In the data mining field, traditional frequent pattern mining is one of the fundamental techniques and has played an important role. Although this approach can find meaningful information based on frequency, it can handle only binary databases and treats items in the databases with the same importance in contrast to many real world applications where there is a need to consider relative significance of items in non-binary databases. Moreover, a threshold setting is required to conduct mining processes in the framework. However, it is difficult for users to set an appropriate threshold since they cannot predict and control mining results precisely. Accordingly, top-k high utility pattern mining has emerged as an essential topic due to its ability to consider the characteristics of real world databases and mine useful information without any threshold setting. In this paper, we analyze relevant methods for top-k high utility pattern mining, evaluate their performance through various experiments, and study their improvements.