Constraint-Based Sequence Mining Using Constraint Programming

Abstract

Abstract. The goal of constraint-based sequence mining is to find sequences of symbols that are included in a large number of input sequences and that satisfy some constraints specified by the user. Many constraints have been proposed in the literature, but a general framework is still missing. We investigate the use of constraint programming as general framework for this task. We first identify four categories of constraints that are applicable to sequence mining. We then propose two constraint programming formulations. The first formulation introduces a new global constraint called exists-embedding. This formulation is the most efficient but does not support one type of constraint. To support such constraints, we develop a second formulation that is more general but incurs more overhead. Both formulations can use the projected database technique used in specialised algorithms. Experiments demonstrate the flexibility towards constraint-based settings and compare the approach to existing methods.

Constraint-Based Sequential Pattern Mining with Decision Diagrams

“…Research in CSPM has primarily focused on exploiting special properties of constraints, such as monotonicity or antimonotonicity, to guarantee the feasibility of pattern extensions in the mining algorithm (Garofalakis, Rastogi, and Shim 1999;Zaki 2000;Lin and Lee 2005;Bonchi and Lucchese 2005;Chen and Hu 2006;Pei, Han, and Wang 2007;Nijssen and Zimmermann 2014;Mallick, Garg, and Grover 2014;Aoga, Guns, and Schaus 2017). Constraint types that do not possess such properties remain a challenge for CSPM algorithms, although some of these have been successfully incorporated in more general item-set mining on databases where events have no specific order (Soulet and Crémilleux 2005;Bistarelli and Bonchi 2007; (C, 2, 1), (C, 5, 2), (A, 8, 3) 2007; Le Bras, Lenca, and Lallich 2009;Leung et al 2012), as well as in CSPM when items and attributes are interchangeable (Pei, Han, and Wang 2007).…”

“…in terms of memory requirements and computational time, in particular when the resulting subset of constrained patterns is small in comparison to the full unconstrained set. Constraint-based sequential pattern mining (CSPM) aims at providing more efficient methods by embedding constraint reasoning within existing mining algorithms (Pei, Han, and Wang 2007;Negrevergne and Guns 2015). Nonetheless, while certain constraint types are relatively easy to incorporate in a mining algorithm, others of practical use are still challenging to handle in a general and effective way.…”

Efficiency Analysis of ASP Encodings for Sequential Pattern Mining Tasks

“…In [Negrevergne and Guns, 2015], the authors organize the constraints on sequential patterns in three categories: 1) constraints on patterns, 2) constraints on patterns embeddings, 3) constraints on pattern sets. These constraints are provided by the user and capture his background knowledge.…”

“…In these experiments, we analyse the proposed encodings on processing real datasets. We use the same real datasets as selected in [Negrevergne and Guns, 2015] to have a representative panel of application domains:…”

“…This version is known to be very efficient, but it cannot cope with embedding constraints, while CPSM-emb does but is less efficient. We do not compare our approach with dedicated algorithms, which are known to be more efficient than declarative mining approaches (see [Negrevergne and Guns, 2015] for such comparisons). The timeout was set to 1 hour.…”