Finding Constrained Frequent Episodes Using Minimal Occurrences

Abstract

Recurrent combinations of events within an event sequence, known as episodes, often reveal useful information.Most of the proposed episode mining algorithms adopt an apriori-like approach that generates candidates and then calculates their support levels. Obviously, such an approach is computationally expensive. Moreover, those algorithms are capable of handling only a limited range of constraints. In this paper, we introduce two mining algorithms -Episode Prefix Tree (EPT) and Position Pairs Set (PPS) -based on a prefix-growth approach to overcome the above limitations. Both algorithms push constraints systematically into the mining process. Performance study shows that the proposed algorithms run considerably faster than MINEPI [4]. * A full version of this paper can be retrieved from

A survey of episode mining

“…Many studies have been done on extracting concise representations of episodes to reduce the number of discovered episodes and have shown that those representations can provide better accuracy for various prediction tasks (Amiri et al, 2019;Ao et al, 2019;Li et al, 2018;Liao et al, 2016Liao et al, , 2018Ma et al, 2004;Méger & Rigotti, 2004;Tatti & Cule, 2011Zhou et al, 2010;Zhu et al, 2012Zhu et al, , 2011.…”

“…Maximal episodes (ME; Ao et al, 2019;Fournier-Viger et al, 2022;Ma et al, 2004) An interesting property of maximal episodes is that ME ⊆ CE ⊆ FE. Thus, maximal episodes are a more compact representation than the closed episodes.…”

“…The second constraint is the maximum amount of time between two consecutive events (gap or inter-event constraint). The consideration of time constraints has been used in WinMiner (Méger & Rigotti, 2004), EPS, and PPT (Ma et al, 2004) among others. Researchers have also studied the integration of the minimum and maximum number of events per episode (length constraint) for the analysis of heterochrony in developmental biology (Bathoorn et al, 2010).…”

“…Besides these three basic episode types, various extensions of episode mining have been proposed. For instance, to reduce the number of episodes presented to users, subsets of episodes that have interesting properties have been studied, such as frequent closed episodes (Ao et al, 2019) and maximal episodes (Ma et al, 2004). Moreover, additional constraints have been added to select more interesting episodes, such as constraints on the time gap between consecutive events (Ma et al, 2004).…”

“…For instance, to reduce the number of episodes presented to users, subsets of episodes that have interesting properties have been studied, such as frequent closed episodes (Ao et al, 2019) and maximal episodes (Ma et al, 2004). Moreover, additional constraints have been added to select more interesting episodes, such as constraints on the time gap between consecutive events (Ma et al, 2004). Algorithms have also been developed to mine episodes in dynamic sequences, and to find the top-k most interesting episodes (Amphawan et al, 2015) instead of those whose frequency is not less than a given threshold.…”