A constructing approach to multi-granularity object-induced three-way concept lattices
Introduction
Formal concept analysis (FCA) was put forward by mathematician Wille in 1982 [1]. This theory provides a concept hierarchy structure based on the binary relation between objects and attributes of datasets. The point of departure for FCA is Boolean data [2], which is named as a formal context in FCA, indicating each object possesses what attributes. The formal concept consists of a set of objects and a set of attributes established by two set-theoretic operators [1]. Concept lattice model has been widely concerned since it was put forward, many scholars [3], [4], [5], [6], [7], [8], [9] extend the original model proposed by Wille. As an effective tool for knowledge characterization and processing, FCA has been widely used in knowledge engineering, machine learning, expert system, information retrieval, data mining, ect [10], [11], [12], [13], [14].
Three-way decision (3WD) is an effective decision-making method with its main idea of “dividing and acting” proposed by Yao in 2012 [15]. It adds a hesitant decision, i.e. delayed decision, on account of the accept-reject two-way decision. The basic opinion of 3WD derives from a three partitions of the universe, including three disjoint regions, i.e. positive, boundary, and negative regions. Rules of accept (reject, delay) decision are extracted from the positive (negative, boundary) region. 3WD has become a theory and method for the fields of granular computing and knowledge discovery together with rough set, fuzzy set, cloud computing and formal concept analysis [16], [17], [18], [19], [20], [21]. In fact, when information is not enough, the non-commitment is usually more in line with the needs of practical problems, therefore the 3WD theory has gradually become an important theory and method to deal with uncertain problems, which has been widely concerned, and has been successfully applied to information, engineering, medical treatment, management and other fields [22], [23], [24].
Previous studies on formal concept analysis were mainly based on positive operators. Missaoui et al. [25] considered both the positive and negative attributes of the formal context, and then performed computing implications. Ganter et al. [26] put forward partial formal contexts, and then researched implicational knowledge. 3WD not only considers positive and negative, but also introduces hesitation. Thus the combination of 3WD and formal concept analysis can lead to richer knowledge based on the formal context. Qi et al. [27] presented three-way concept analysis, and proposed two types of three-way concept lattice. After that, Qi and Mao et al. further studied connections among three-way concept and classical concept lattices [28], [29]. Lots of scholars conducted extended research on the three-way concept lattices [30], [31], [32]. Zhao et al. [33] investigated the connections among the current several three-way concept lattice models. Wei et al. [34] defined decision rules of formal decision contexts and proposed the rules acquisition method based on three-way concept lattices. Qian et al. [35] proposed a construction algorithm by two new types of formal contexts whose concept lattices and three-way concept lattice are isomorphic, and showed that their algorithm outperforms the algorithm of constructing the OE-concept lattice by its definition. Yang et al. [36] proposed a new construction algorithm via the concepts of a formal context and its complement context, and expressed that their method outperforms existing methods. Yao [37] systematically analyzed and summarized FCA based on 3WD perspectives. Qi et al. [38] proposed the notion of 3-valued formal contexts, defined 3-valued operators and constructed 3-valued concept lattices. The relationships between 3-valued lattices and existing approximation concept lattices are examined.
Lin [39] proposed granular computing theory in 1998, this idea was derived from fuzzy information granulation put forward by Zadeh [40]. Its early results mainly focused on granulation principles, granulation methods, and granulation normal form research [41], [42], [43], [44]. Pawlak [45] presented rough set theory in 1982, which is an effective mathematical tool for dealing with incomplete and imprecise information. The classic rough set theory is based on an equivalence relation (granularity). However, single-granular rough sets cannot be used to describe the multi-perspective idea in granular computing theory. Therefore Wu et al. [46] proposed a multi-scale information system from the perspective of granular computation which has different levels of granulations. In such a system, each object under each attribute is represented by different scales at different levels of granulations having a granular information transformation from a finer to a coarser labelled value. In recent years, inspired by multi-scale information system, some scholars have introduced multi-scale ideas into FCA and proposed multi-granularity (labelled) formal context. In fact, the idea of multi-scale in FCA has been reflected in the literature [47]. Constructing concept lattices is one of the most important problem of FCA. Therefore, the construction of concept lattices based on the multi-granularity formal context is a research hotspot. From the perspective of algorithm, the zoom algorithms for classical concept lattice and object (property) oriented concept lattice are given through a granularity tree proposed by Belohlavek and Shao et al. [47], [48], [49], respectively. Based on the perspective of the transformation relationship among concepts, Hu and Qin [50] proposed constructing methods for classical concept lattice under a multi-granularity formal context.
In real life, if the attribute granularity level is too fine, lots of redundant concepts may be generated, thus it takes a lot of time to extract useful information. On the other hand, if the attribute granularity level is too coarse, some useful knowledge may be ignored. Therefore, different application situations may require different granularity level. When the granularity level changes, it takes a lot of time to reconstruct the three-way concept lattice by the algorithms proposed in [36], because the concept lattice of the original formal context is ignored. Compared with the classical concept lattice construction, there are not researches about constructing object-induced three-way concept lattices based on a multi-granularity formal context. Therefore, this paper mainly researches a constructing approach to object-induced three-way multi-granularity concept lattices.
The structure of this article is as follows: FCA and three-way concept lattices are concisely reviewed in section 2. Section 3 investigates connections among object-induced three-way concepts under a multi-granularity formal context and the construction algorithms for OE-concept lattice based on a multi-granularity formal context. In section 4, we compare these algorithms with the recent construction algorithms [36] for the three-way concept lattice generation to show the superiority of these proposed algorithms. In section 5, the summary of this article and the future outlook are drawn.
Section snippets
Preliminaries
In this section, the related theoretical foundations of FCA and three-way concept lattice are briefly reviewed. Please refer to [1], [27] for details.
Constructing multi-granularity OE-concept lattices
In a Pawlak information system, each object only possesses one value for each attribute. But, in some real datasets, each object under each attribute may be represented by different scales at different levels of granularities. Therefore, from the perspective of granular computing, Wu et al. [46] introduced multi-scale information table which has different levels of granularities. Inspired by these, Hu et al. [50] proposed the notion of multi-granularity formal context to deal with
Experimental analysis
In the previous section, we put forward the construction algorithms for OE-concept lattices based on a multi-granularity formal context. In this section, we carry out experiments to evaluate the performance of our proposed algorithms by comparing with the latest static construction algorithms presented by Yang et al. [36].
Conclusion
Multi-granularity and three-way concept lattice have become research hotspots in recent years. We mainly focus on a constructing approach to multi-granularity OE-concept lattices, present the mutual transformation methods of OE-concepts lattices under different granularity. Then, the applicability of these methods is shown by some examples and the corresponding algorithms are presented. Finally, several datasets from UCI are selected to illustrate the effectiveness of our algorithms, and the
CRediT authorship contribution statement
Qian Hu: Conceptualization, Methodology, Software, Writing – original draft. Keyun Qin: Methodology, Validation, Writing – review & editing. Lei Yang: Software, Validation, Writing – review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgement
This work has been partially supported by the National Natural Science Foundation of China (Grant No. 61976130).
References (51)
Concept lattices and order in fuzzy logic
Ann. Pure Appl. Log.
(2004)- et al.
Variable threshold concept lattices
Inf. Sci.
(2007) - et al.
A multiview approach for intelligent data analysis based on data operators
Inf. Sci.
(2008) - et al.
On heterogeneous formal contexts
Fuzzy Sets Syst.
(2014) Semantic web search based on rough sets and fuzzy formal concept analysis
Knowl.-Based Syst.
(2012)Three-way decisions with probabilistic rough sets
Inf. Sci.
(2010)- et al.
Systematic studies on three-way decisions with interval-valued decision-theoretic rough sets
Inf. Sci.
(2014) The superiority of three-way decisions in probabilistic rough set models
Inf. Sci.
(2011)- et al.
A three-way decision method based on fuzzy rough set models under incomplete environments
Inf. Sci.
(2021) - et al.
Preference degree-based multi-granularity sequential three-way group conflict decisions approach to the integration of tcm and western medicine
Comput. Ind. Eng.
(2020)
The connections between three-way and classical concept lattices
Knowl.-Based Syst.
The attribute reductions of three-way concept lattices
Knowl.-Based Syst.
Three-way dual concept analysis
Int. J. Approx. Reason.
On relationship between three-way concept lattices
Inf. Sci.
Rules acquisition of formal decision contexts based on three-way concept lattices
Inf. Sci.
Constructing three-way concept lattices based on apposition and subposition of formal contexts
Knowl.-Based Syst.
Constructing three-way concept lattice based on the composite of classical lattices
Int. J. Approx. Reason.
Three-way granular computing, rough sets, and formal concept analysis
Int. J. Approx. Reason.
Toward a theory of fuzzy information granulation and its centrality in human reasoning and fuzzy logic
Fuzzy Sets Syst.
Granular computing and dual Galois connection
Inf. Sci., Int. J.
Optimal scale selection for multi-scale decision tables
Int. J. Comput. Inf. Sci.
Granularity of attributes in formal concept analysis
Inf. Sci.
Restructuring lattice theory: an approach based on hierarchies of concepts
Formal Concept Analysis
The study of the l-fuzzy concept lattice
Mathw. Soft Comput.
Cited by (10)
BTWM-HF: A behavioral three-way multi-attribute decision-making method with hesitant fuzzy information
2024, Expert Systems with ApplicationsGraph representation learning method based on three-way partial order structure
2024, International Journal of Approximate ReasoningAn efficient method of renewing object-induced three-way concept lattices involving decreasing attribute-granularity levels
2024, International Journal of Approximate ReasoningUncertainty and three-way decision in data science
2023, International Journal of Approximate ReasoningSequential 3WD-based local optimal scale selection in dynamic multi-scale decision information systems
2023, International Journal of Approximate ReasoningCitation Excerpt :Zhang et al. [28] used sequential 3WD model to construct multi-granularity auto encoder features. Hu et al. [29] used the sequential three-way decision making method to study the construction method of three-way concept lattices induced by objects in the multi-granularity formal contexts. Zhao et al. [30] applied sequential 3WD to the multi-label classification, and constructed a trisecting-acting-outcome framework.
Formal concept analysis perspectives on three-way conflict analysis
2023, International Journal of Approximate ReasoningCitation Excerpt :One of its main ideas is to divide a whole into three parts and to apply strategies to act upon the three parts for achieving a desirable outcome. There are extensive studies on the development of three-way decision and the combinations of it with other theories, for example, game-theoretic three-way decision [1,59], three-way learning algorithms [2,19,37,42,57,60,62], three-way decision with rough sets [9,22,36,48,56], three-way formal concept analysis [8,14–16,27,28,30,35,39,41,45,51,53,61,63,65], three-way approximations of fuzzy sets [18,52,58,66], three-way multi-attribute decision-making [20,38,54], and many more. Three-way conflict analysis is one of the most recent research efforts [5,11–13,17,21,23,34,47,50,55,64].