Elsevier

Knowledge-Based Systems

Volume 91, January 2016, Pages 165-178
Knowledge-Based Systems

Approximate concept construction with three-way decisions and attribute reduction in incomplete contexts

https://doi.org/10.1016/j.knosys.2015.10.010Get rights and content

Abstract

Incomplete contexts are a kind of formal contexts in which the relationship between some objects and some attributes is unavailable or lost. Knowledge discovery in incomplete contexts is of interest because such databases are frequently encountered in the real world. This paper mainly focuses on two issues: approximate concept construction with three-way decisions and attribute reduction in incomplete contexts. The theory of three-way decisions is formulated based on the notions of acceptance, rejection and non-commitment. It is an extension of the commonly used binary-decision model with an added third option. Based on three-way decisions, we propose two models to construct approximate concepts in incomplete contexts, and the equivalence of the two is revealed. To simplify the representation of the approximate concept lattices, we further present the attribute reduction approaches.

Introduction

The theory of formal concept analysis (FCA), proposed by Wille [14], [63], is a mathematical tool for conceptual data analysis and knowledge processing. It mainly concerns the lattice structure, called a concept lattice, constructed by the formal concepts embedded in a concept hierarchy. A formal concept, is a pair consisting of a set of objects (the extension), which includes all objects belonging to the concept, and a set of attributes (the intension), which includes all the attributes common to objects in the extension. In the last two decades, FCA has been successfully applied to information retrieval, knowledge discovery, data mining, machine learning, software engineering, etc. [2], [4], [6], [10], [13], [24], [25], [46], [47], [48], [56], [57].

As is well known, knowledge reduction is a key issue in rough set theory [45], and the main objective is to simplify the representation of knowledge so as to gain optimized decision rules. Recently a large number of reduction methods have been developed for different types of information systems and decision tables [8], [19], [22], [42], [51], [62], [67]. Similarly, knowledge reduction in FCA has also received much attention in the past decades [5], [9], [14], [16], [26]. Specially, much attention has been paid to attribute reduction of formal contexts in FCA. For example, Ganter and Wille [14] proposed to reduce the context by removing both the reducible attributes and objects. The concept lattice obtained from the reduced context is isomorphic to the one induced from the original context. Zhang et al. [77] presented an approach to attribute reduction by keeping all the extensions unchanged. Wang and Zhang [60], Li et al. [34] conducted attribute reduction based on irreducible elements respectively, which are equivalent to Zhang et al.’s method without generating all the concepts. Wu et al. [65] removed redundant attributes to keep information granule, i.e., all extensions of the object concepts, unchanged. Shao and Leung [53] revealed that granular reducts obtained by Wu et al.’s approach [65] are equivalent to dominance reducts. Some researchers focus on the issue of attribute reduction from other perspectives instead of classical FCA. For example, Liu et al. [39] conducted attribute reduction in rough set-based concept lattice [12], [68], Li and Wu [35] reduced attributes from the perspective of covering rough set, Mi et al. [41] gave the approaches to attribute reduction in concept lattices induced by axialities. Besides, some attribute reduction approaches in decision formal context [61] are proposed based on rule acquisition [29], [30], [33], [54].

There is an underlying assumption in classical concept lattice and the extended versions: the binary relation in the formal context is definite, and either the object x has the attribute a, or the opposite holds. In many real-world situations, however, it may happen that the relationship between some objects and some attributes is unavailable or lost, and there may be a third relation between the object and the attribute. In this case, the contexts are called incomplete contexts [3]. (And the classical formal contexts are called complete ones.)

Recently, there is a growing interest in the study of the incomplete contexts [3], [11], [18], [23], [31], [44]. However, most of the research focuses on attribute implication or attribute exploration, and few efforts have been devoted to concept construction in incomplete contexts. Although Li et al. [31] have given an approximate concept construction approach by representing each concept intension as a pair (definite common attributes, possible common attributes), they only take the positive information, i.e., the common (or possible common) attributes shared by the extension, into consideration. Nevertheless, in some cases, some negative information is of equal importance. For example, when a patient chooses some Over-The-Counter (OTC) medicines for a special disease, he must pay attention to the indications (the positive information) and the contraindications (the negative information) of the medicines. In this case, the objects are the OTC medicines, and the attributes are diseases. The indications, the contraindications, and other diseases that are not listed in the medicines’ instruction form a tripartition of the attribute set, which falls into the category of three-way decisions.

The theory of three-way decisions, proposed by Yao [73], is an extension of the commonly used acceptance-rejection binary model with an added third option: non-commitment [70], [71], [72], [73], [74]. The basic ideas in the theory of three-way decisions are based on a tripartition of a universal set. A tripartition consists of three pair-wise disjoint subsets, namely the positive region POS, the negative region NEG and the boundary region BND, of the universe set [73]. Corresponding to the three regions, one can construct rules for acceptance, rejection and non-commitment, respectively [70], [71]. Three-way decisions play a key role in everyday decision-making and have been widely used in many fields and disciplines [1], [17], [20], [21], [28], [32], [36], [37], [38], [40], [58], [66], [74], [75], [76], [78].

In fact, some researchers have introduced the theory of three-way decisions into FCA in complete contexts. For example, Ganter and Kuznetsov [15] pointed out that the version space is by definition the set of classifiers that match all positive examples and do not match any negative examples. Rodriguez-Jimenez et al. proposed the mixed formal concept lattice, in which both positive and negative attributes are contained, to extract mix rules [52]. Qi et al. proposed two three-way formal concept lattices, namely the OE-concept lattice and the AE-concept lattice [49].

Motivated by the works mentioned above, this paper focuses on approximate concept construction with three-way decisions and attribute reduction in incomplete contexts. Compared with other methods, for example [31], that deal with missing information, the advantage of using three-way decisions is that it deals with the missing information from two aspects simultaneously, the positive aspect and the negative aspect.

The rest of the paper is organized as follows. Section 2 briefly reviews some basic notions related to complete contexts and incomplete contexts. Section 3 proposes two models to construct three-way approximate concepts in incomplete contexts, and then the equivalence of the two models is proved. After that, Section 4 discusses attribute reduction and attribute characteristics in the three-way approximate concept lattices. Finally, Section 5 concludes this paper.

Section snippets

Preliminaries

In this section, we briefly review some basic notions related to complete contexts, and then introduce the notion of incomplete context.

Definition 1

[14]

A formal context C=(U,A,R) consists of two sets U and A and a binary relation R between U and A. The elements of U are called objects, and the elements of A are called attributes. For an object x and an attribute a, (x, a) ∈ R indicates the object x has the attribute a, and (x,a)R indicates the opposite.

A formal context can be represented by a two-dimensional

Approximate concept construction with three-way decisions in incomplete contexts

Just as mentioned in the introduction, the positive information (obtained by the positive operators) is as important as the negative information (obtained by the negative operators) to describe a concept. To achieve this goal, in this section, we propose two models based on three-way decisions to construct the approximate concept lattice of an incomplete context, one is based on three-way approximate operators, and the other is based on the tripartition of the ternary relation.

Attribute reduction and attribute characteristics in three-way approximate concept lattices

In this section, we mainly focus on the issues of attribute reduction and attribute characteristics in three-way approximate concept lattices. First, we recall some basic notions and existing approaches to the two issues related to the classical concept lattice.

Conclusions

In this paper, we proposed two models to construct approximate concept with three-way decisions in incomplete contexts. One is from the perspective of the three-way approximate operators, and the other is from perspective of the tripartition of the ternary relation of the incomplete context. In each model, the approximate concepts contains both the positive and the negative information simultaneously. It has been proved that the two models are equivalent to each other. After the approximate

Acknowledgment

The authors would like to thank the Editor and the anonymous reviewers for their valuable comments and constructive suggestions. This work is supported by the National Natural Science Foundation of China (Nos. 61272060 and61379114), and Key Natural Science Foundation of Chongqing (No. CSTC2013jjB40003), and the Social Science Foundation of the Chinese Education Commission (No. 15XJA630003).

References (78)

  • J.H. Li et al.

    Incomplete decision contexts: Approximate concept construction, rule acquisition and knowledge reduction

    Int. J. Approx. Reason.

    (2013)
  • J.L. Li et al.

    Multistage email spam filtering based on three-way decisions

    Rough Sets and Knowledge Technology

    (2013)
  • J.S. Mi et al.

    Approaches to attribute reduction in concept lattices induced by axialities

    Knowledge-Based Syst.

    (2010)
  • J.S. Mi et al.

    Approaches to knowledge reduction based on variable precision rough set model

    Inf. Sci.

    (2004)
  • E.M. Norris

    An algorithm for computing the maximal rectangles in a binary relation

    Revue Roumaine de Mathématiques Pures et Appliquées

    (1978)
  • S. Obiedkov

    Modal logic for evaluating formulas in incomplete contexts

    Conceptual Structures: Integration and Interfaces

    (2002)
  • U. Priss

    Formal concept analysis in information science

    Ann. Rev. Inf. Sci. Technol.

    (2006)
  • M.W. Shao et al.

    Rule acquisition and complexity reduction in formal decision contexts

    Int. J. Approx. Reason.

    (2014)
  • A. Skowron et al.

    The discernibility matrices and functions in information systems

    Intelligent Decision Support

    (1992)
  • G. Stumme et al.

    Conceptual Knowledge Discovery in Databases using Formal Concept Analysis Methods

    (1998)
  • T. Tilley et al.

    A survey of formal concept analysis support for software engineering activities

    Formal Concept Analysis

    (2005)
  • L. Wei et al.

    Attribute reduction theory of concept lattice based on decision formal contexts

    Sci. China Ser. F Inf. Sci.

    (2008)
  • Y.Y. Yao

    Three-way decision: an interpretation of rules in rough set theory

    Rough Sets and Knowledge Technology

    (2009)
  • Y.Y. Yao

    Three-way decisions using rough sets

    Rough Sets: Selected Methods and Applications in Management and Engineering

    (2012)
  • H. Yu et al.

    A cluster ensemble framework based on three-way decisions

    Rough Sets and Knowledge Technology

    (2013)
  • N. Azam et al.

    Multiple criteria decision analysis with game-theoretic rough sets

    Rough Sets and Knowledge Technology

    (2012)
  • P. Burmeister et al.

    On the treatment of incomplete knowledge in formal concept analysis

    Conceptual Structures: Logical, Linguistic, and Computational Issues

    (2000)
  • C. Carpineto et al.

    Exploiting the potential of concept lattices for information retrieval with credo

    J. Univers. Comput. Sci.

    (2004)
  • R. Cole et al.

    Document retrieval for e-mail search and discovery using formal concept analysis

    Appl. Artif. Intell.

    (2003)
  • B.A. Davey et al.

    Introduction to Lattices and Order

    (2002)
  • D.G. Chen et al.

    A new approach to attribute reduction of consistent and inconsistent covering decision systems with covering rough sets

    Inf. Sci.

    (2007)
  • D. Dubois et al.

    A possibility-theoretic view of formal concept analysis

    Fundamenta Informaticae

    (2007)
  • I. Düntsch et al.

    Modal-style operators in qualitative data analysis

    Proceedings of 2002 IEEE International Conference on Data Mining (ICDM 2002)

    (2002)
  • D.M. Endres et al.

    An application of formal concept analysis to semantic neural decoding

    Ann. Math. Artif. Intell.

    (2009)
  • B. Ganter et al.

    Formal Concept Analysis: Mathematical Foundations

    (1999)
  • B. Ganter et al.

    Hypotheses and Version Spaces, Conceptual Structures for Knowledge Creation and Communication, Formal concept analysis: Mathematical Foundations

    (2003)
  • J.P. Herbert et al.

    Game-theoretic rough sets

    Fundamenta Informaticae

    (2011)
  • R. Holzer

    Knowledge acquisition under incomplete knowledge using methods from formal concept analysis: part i and part ii

    Fundamenta Informaticae

    (2004)
  • M. Inuiguchi et al.

    Variable-precision dominance-based rough set approach and attribute reduction

    Int. J. Approx. Reason.

    (2009)
  • Cited by (109)

    • Formal concept analysis perspectives on three-way conflict analysis

      2023, International Journal of Approximate Reasoning
    View all citing articles on Scopus
    View full text