Selection of Satisfied Association Rules via Aggregation of Linguistic Satisfied Degrees

Many association rule mining algorithms have been well-established, such as Apriori, Eclat, FP-Growth, or LCM algorithms. However, the challenge is that the huge size of association rules is extracted by using these algorithms, and it is difficult for users to select satisfied association rules from them. In this paper, a new method is proposed to select satisfied association rules, which is based on the aggregation of fuzzy linguistic satisfied degrees of extracted association rules. To this end, two problems must be solved, one is which interesting measures are utilized to obtain fuzzy linguistic satisfied degrees of association rules and the other is how to aggregate them. For the first problem, many objective and subjective interesting measures have been proposed, which are generally included in [0, 1] or others universes and easily calculated by support, confidence, or other measures, these interesting measures cannot be directly aggregated, because different interesting measures represent different satisfied degrees of association rules. In this paper, a new transformation function is proposed to transform these interesting measures into fuzzy linguistic satisfied degrees, such as dissatisfied, fair, satisfied, and so on. For the second problem, by considering different weights of objective and subjective interesting measures, linguistic aggregation operators are designed to aggregate these fuzzy linguistic satisfied degrees of association rules. Accordingly, satisfied association rules are selected by using order on the aggregation results of linguistic satisfied degrees. In cases' study, Apriori, Eclat, FP-Growth, and LCM algorithms are first utilized to extract association rules with higher support or confidence measures from Chess, Connect, Mushroom, and T40I10D100K databases, then the proposed method is applied to obtain fuzzy linguistic satisfied degrees of extracted association rules and aggregate and select satisfied association rules from those extracted association rules, and comparison and analysis show that the proposed method is a useful and alternative tool to select satisfied association rules from extracted association rules.