Managing the Absence of Items in Fuzzy Association Mining

One of the most well-known and extended data mining techniques is that of association rule mining, a helpful tool to discover relations between items present in sets of transactions. Nevertheless, in some other scenarios, another interesting issue is that of considering not only the possible relations involving presence of items, but the absence of them. The problem gets more complex when it is necessary to represent also imprecision and/or uncertainty in the information. In this paper, we introduce a methodology to obtain fuzzy association rules involving absent items. Additionally, our proposal is based on restriction level sets, a recent representation of imprecision that extends that of fuzzy sets, and introduces some new operators, covering some misleading results obtained from usual fuzzy operators as, for example, negation. In our methodology, we define new measures of interest and accuracy for fuzzy association rules as RL-numbers, as well as we propose a new way of summarizing the resulting set of fuzzy association rules, distributed in restriction levels. Keywords—Absence of items, fuzzy association rules, restriction levels