The pattern mining problem has derived to more user-centric techniques to extract not just frequent itemsets but patterns or set of items that provide much richer insights [Lun20]. An example of this is the task of mining high utility patterns [Chu15], that is, those having weights indicating their relative importance to the user. This emerging topic has been studied for about a decade and has attracted the attention of many researchers. The goal is to identify all sequences or patterns having a high utility (importance) in a database. To represent the usefulness of patterns, each item may be associated with a unit profit indicating its relative importance, and each occurrence of the item is associated with a quantity.
High utility pattern mining has greatly evolved since it was presented a decade ago, and the task is now more user centric. There are studies [Fou19] in which both internal and external utilities are considered. The internal utility of an item denotes its importance in a transaction or data record, e. g. the quantity of the item. The external utility, on the contrary, describes user preferences and highly depends on the utility function specified by the user. Some other studies have addressed the problem of reducing the resulting set of solutions to improve the users’ understanding. In this sense, some studies [Din18] have considered the periodic appearance of patterns as a criterion to select promising solutions. Thus, the idea is not only to maximize the utility but to produce cyclic pieces of evidence of the extracted patterns. However, even when the high utility pattern mining task is widely accepted as a research line, it is still extendable with novel techniques that enable to improve its actionability.
The high utility task has been widely studied and efficient algorithms has been proposed up to date, but this task can be improved in different ways:
- Development of algorithms for mining high utility patterns that improve the performance of existing algorithms in terms of computational time and memory consumptions. The algorithms should be able to constraint the solutions according to the users’ aim, so the resulting set is according to the users’ expectations.
- Design of new high utility mining algorithms in which subjective knowledge, based on the users’ experience, is introduced.
- Development of novel algorithms for mining high utility patterns considering the periodicity of the records. The aim is to work on continuous data without any prior discretization step and, therefore, minimizing the loss of information.
The research group has an important background in the pattern mining field [Ven16] and, recently, has developed some proposals to extract high utility patterns [Fou20] as well as local periodic patterns [Fou21] as well as user-centric knowledge [Lun20].
[Ven16] S. Ventura, J. M. Luna: Pattern Mining with Evolutionary Algorithms. Springer 2016, ISBN 978-3-319-33857-6, pp. 1-190.
[Fou21] P. Fournier-Viger, P. Yang, R. U, Kiran, S. Ventura, José M. Luna: Mining local periodic patterns in a discrete sequence. Information Science. 544: 519-548 (2021).
[Lun20] J. M. Luna, P. Fournier-Viger, S. Ventura: Extracting User-Centric Knowledge on Two Different Spaces: Concepts and Records. IEEE Access 8: 134782-134799 (2020).
[Fou20] P. Fournier-Viger, Y. Wang, J. Chun-Wei Lin, J. M. Luna, S. Ventura: Mining Cross-Level High Utility Itemsets. IEA/AIE 2020: 858-871.
[Chu15] J. Chun-Wei Lin, W. Gan, T. P. Hong, V. S. Tseng: Efficient algorithms for mining up-to-date high-utility patterns. Adv. Eng. Informatics 29(3): 648-661 (2015)
[Din18] D. T. Dinh, B. Le, P. Fournier-Viger, V. N. Huynh: An efficient algorithm for mining periodic high-utility sequential patterns. Appl. Intell. 48(12): 4694-4714 (2018)
[Fou19] P. Fournier-Viger, J. C. W. Lin, R. Nkambou, B. Vo, V. Tseng: High-Utility Pattern Mining: Theory, Algorithms and Applications. Springer 2019. ISBN 978-3-030-04920-1