Frequent itemset mining: A 25 years review

Abstract Frequent itemset mining (FIM) is an essential task within data analysis since it is responsible for extracting frequently occurring events, patterns, or items in data. Insights from such pattern analysis offer important benefits in decision‐making processes. However, algorithmic solutions for mining such kind of patterns are not straightforward since the computational complexity exponentially increases with the number of items in data. This issue, together with the significant memory consumption that is present in the mining process, makes it necessary to propose extremely efficient solutions. Since the FIM problem was first described in the early 1990s, multiple solutions have been proposed by considering centralized systems as well as parallel (shared or nonshared memory) architectures. Solutions can also be divided into exhaustive search and nonexhaustive search models. Many of such approaches are extensions of other solutions and it is therefore necessary to analyze how this task has been considered during the last decades. This article is categorized under: Algorithmic Development > Association Rules Technologies > Association Rules