In the context of pattern mining, the utility of a pattern can be described as a preference ordering over a choice set; it can be actually assessed from very different perspectives and at different abstraction levels. However, while the topic of High-Utility Pattern Mining (HUPM) has been widely studied, the basic assumption is that each item in a knowledge base is associated with one, static utility. In this paper we introduce, among others, the notion of facets for items, which allows to cope with this limitation and, moreover, we show how a more structured representation of available information, coupled with facets defined also for higher abstraction levels, paves the way to new opportunities for HUPM. In particular, the proposed framework allows to introduce some new advanced classes of utility functions in the detection process, whose relevance is also experimentally evaluated. A real use case on paper reviews is exploited to analyze the potentiality of the proposed framework in knowledge creation and discovery. Given the wide variety of analytical scenarios that can be envisioned in this new setting, we take full advantage of the capabilities of Answer Set Programming and its extensions for a fast encoding and testing of the framework.
An Answer Set Programming Based Framework for High-Utility Pattern Mining Extended with Facets and Advanced Utility Functions
Cauteruccio F.;
2021-01-01
Abstract
In the context of pattern mining, the utility of a pattern can be described as a preference ordering over a choice set; it can be actually assessed from very different perspectives and at different abstraction levels. However, while the topic of High-Utility Pattern Mining (HUPM) has been widely studied, the basic assumption is that each item in a knowledge base is associated with one, static utility. In this paper we introduce, among others, the notion of facets for items, which allows to cope with this limitation and, moreover, we show how a more structured representation of available information, coupled with facets defined also for higher abstraction levels, paves the way to new opportunities for HUPM. In particular, the proposed framework allows to introduce some new advanced classes of utility functions in the detection process, whose relevance is also experimentally evaluated. A real use case on paper reviews is exploited to analyze the potentiality of the proposed framework in knowledge creation and discovery. Given the wide variety of analytical scenarios that can be envisioned in this new setting, we take full advantage of the capabilities of Answer Set Programming and its extensions for a fast encoding and testing of the framework.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.