Mining Gradual Itemsets Using Sequential Pattern Mining
Saïd Jabbour, Jerry Lonlac, Lakhdar Saïs · 2019
Gradual itemsets model complex attributes covariation of the form "The more or less is A, the more or less is B". Recently, such kind of itemsets have received attention from the data mining community, where several formalizations and methods have been defined to automatically extract and maintain gradual patterns from numerical databases. However, mining gradual itemsets remains challenging as the task is more complex than ordering the transactions according to several dimensions or attributes. In fact, the order in which attributes are considered impacts the sorting operation. One can note that an ordering of the transactions according to a single attribute leads to a sequence of itemsets where items correspond to transaction identifiers. In this paper and from this observation, we propose a new formulation of the gradual itemset mining task as the problem of sequential pattern mining. This original reduction allows us to exploit sequential pattern mining algorithms to extract gradual itemsets. Experimental results obtained on several numerical datasets show the feasibility of our proposed framework.