key: cord-0052339-u3ugo8t5
authors: El Atik, Abd El Fattah; Nawar, Ashraf; Atef, Mohammed
title: Rough approximation models via graphs based on neighborhood systems
date: 2020-11-05
journal: Granul
DOI: 10.1007/s41066-020-00245-z
sha: 57610100da322104e2a30015178b35a06a953d55
doc_id: 52339
cord_uid: u3ugo8t5

Neighborhood systems are used to approximate graphs as finite topological structures. Throughout this article, we construct new types of eight neighborhoods for vertices of an arbitrary graph, say, j-adhesion neighborhoods. Both notions of Allam et al. and Yao will be extended via j-adhesion neighborhoods. We investigate new types of j-lower approximations and j-upper approximations for any subgraph of a given graph. Then, the accuracy of these approximations will be calculated. Moreover, a comparison between accuracy measures and boundary regions for different kinds of approximations will be discussed. To generate j-adhesion neighborhoods and rough sets on graphs, some algorithms will be introduced. Finally, a sample of a chemical example for Walczak will be introduced to illustrate our proposed methods.

Motivated by many analyzes requiring rough sets, the present paper aims for a new approach to the study of rough sets from the points of view of both neighborhood systems and graphs. Neighborhood systems on graphs based on rough sets are a generalization of Pawlak's rough set model.

Rough set theory was initially developed (Pawlak 1981 ) as a new mathematical methodology to deal with the vagueness and uncertainty in information systems. Many proposals made for generalizing and interpreting rough sets (Orlowska and Pawlak 1984; Pomykala 1987; Skowron and Stepaniuk 1996; Yao and Line 1996; Zirako 1994 (Allam et al. 2005; Benouini et al. 2020; Dong et al. 2004; Jensen and Shen 2004; Leung et al. 2006; Pal and Mitra 2004; Yao and Chen 2005; Zhao and Liu 2011; Zhan et al. 2019) ]. In 1999, Yao (1999) introduced generalized rough sets through a binary relation; while, these approximations are not satisfied with Pawlak's properties that were applied on an equivalence relation. For this reason, Zhu (2007) studied rough approximations that depend on general relations. These approximations help to prove some properties that were not easy to prove in the classical case. From this time onwards, many types of approximations are investigated. In 2008, Abu-Donia (2008) discussed three types of lower approximations and upper approximations with respect to any binary relation based on the right neighborhoods. This generalization of approximations converted into two ways via a finite number of binary relations. In 2014, Abd El- Monsef et al. (2015) presented the main ideas about the concept of j-neighborhood systems and studied eight approaches for approximating rough sets. Many researchers studied the j-neighborhood systems on different spaces such as Abbas et al. (2016) ; Amer et al. (2017) ; Atef et al. (2020) ; Hosny (2018) ; Huang and Li (2018) ; Kozae et al. (2019) .

Graph theory (Chartrand et al. 2016 ) is an important mathematical tool in diverse subjects. A graph G ¼ ðV; EÞ, is an ordered pair of different sets (V, E), where V is a nonempty set and E is a subset of unordered pairs of V. The vertices and edges of a graph G are the elements V ¼ VðGÞ and E ¼ EðGÞ, respectively. A graph G is finite (respectively, infinite) if the set V(G) is finite (respectively, infinite). The degree of a vertex u 2 VðGÞ is the number of edges containing u. If there is no edge in a graph G but contains a vertex u, then u is called an isolated point, and so the degree of u is zero. An edge that has the same vertex to end is called a loop, and the edge with a distinct end is called a link. A graph is simple if it has no loop and no pair of its links join the same pair of vertices. A graph that has no edge is called a null graph. A directed graph is a graph in which edges have a certain way. In addition, an undirected graph is a graph in which edges have no way. Many scholars work on the theory of graphs and applied it in many fields, see Atef et al. 2020; Liu et al. 2020; Qin et al. 2018; Mandal and Ranadive 2019; William-West and Singh 2018) . In 2018, Nada et al. (2018) initiated the study on topological structures via graphs based on the right neighborhoods. Recently, the neighborhood systems, rough sets on graphs are used to represent structures such as self-similar fractals (El Atik and Nasef 2020) and human heart (El Atik and Nasef 2020) which are useful in physics and medicine, respectively.

As a continuation of the development in the use of general relations, we construct in the present paper new types of j-adhesion neighborhoods from adjacent vertices of general graphs. Based on j-adhesion neighborhoods, we define j-lower approximations and j-upper approximations and the comparison between them and some other types of lower approximations and upper approximations will be discussed. In Sect. 2, we present the fundamental concepts and properties of that used in this paper. In Sect. 3, we introduce the new concepts of j-adhesion neighborhoods and study their basic properties and examples. The goal of Sect. 4 is to generalize some of Pawlak's properties. A comparison between the proposed method and the previous one is shown in Sect. 5. Finally, we apply the results on a sample that is deduced from a reduction by similarity (El Atik 2020) for Walczak's example in chemistry.

In this section, some basic notions of rough sets, graph theory, and a j-neighborhood system will be presented.

Definition 1 (Yao 1999 ) Let R be a binary relation on a nonempty set U and A U. Lower approximations and upper approximations of A are defined by RðAÞ ¼ fx 2 U : xR Ag; and RðAÞ ¼ fx 2 U : xR \ A 6 ¼ /g , where xR ¼ fy 2 U : xRyg.

The following properties of lower approximations and upper approximations for Pawlak (Pawlak 1982; Pawlak and Skowron 1994; Pawlak 1997 ) will be stated.

Definition 2 (Abu-Donia and Salama 2012) Let R be a binary relation on U and A U. Then, the following properties are held.

(1) Roughly R-definable, if RðAÞ 6 ¼ / and RðAÞ 6 ¼ X;

(2) Internally R-undefinable, if RðAÞ ¼ / and RðAÞ 6 ¼ X;

(3) Externally R-undefinable, if RðAÞ 6 ¼ / and RðAÞ = X; (4) Totally R-undefinable, if RðAÞ ¼ / and RðAÞ ¼ X. (1) N r ðxÞ ¼ fy 2 VðGÞ : xRyg.

(2) N l ðxÞ ¼ fy 2 VðGÞ : yRxg.

3 Rough approximation model via graphs using j-neighborhood systems

In this section, we extent Definition 1 of Yao in terms of jneighborhood systems which are defined in Definition 4. We first give an example to illustrate j-neighborhood systems from a simple graph.

Example 1 Let G be a simple graph as shown in Fig. 1 . jneighborhood systems are defined as follows:

Definition 5 Let G ¼ ðVðGÞ; EðGÞÞ be a graph and H be a subgraph of G. Define a first type of lower approximations and upper approximations of H which are denoted by N j ðVðHÞÞ and N j ðVðHÞÞ, respectively.

N j ðVðHÞÞ ¼ fx 2 VðGÞ : N j ðxÞ VðHÞg; and N j ðVðHÞÞ ¼ VðHÞ S fx 2 VðGÞ : N j ðxÞ \ VðHÞ 6 ¼ /g.

If j ¼ r in Definition 5, then we have approximations in Definition 1.

) be a graph and H be a subgraph of G. Define the j-boundary, j-positive, j-negative regions and j-accuracy measure of H in terms of j-adhesion neighborhood which will be denoted by BND P j , POS P j , NEG P j and a N j , respectively.

(1) BND N j ðVðHÞÞ ¼ N j ðVðHÞÞ À N j ðVðHÞÞ.

(2) POS N j ðVðHÞÞ ¼ N j ðVðHÞÞ.

(3) NEG N j ðVðHÞÞ ¼ VðGÞ À N j ðVðHÞÞ.

the results are also by the same manner.

Theorem 1 Let G ¼ ðVðGÞ; EðGÞÞ be a graph and H and K be subgraphs of G. Then, the following properties are held.

(1) N j ðVðGÞÞ ¼ VðGÞ.

(2) If VðHÞ VðKÞ, then N j ðVðHÞÞ N j ðVðKÞÞ.

Proof It is sufficient to prove (1), (2), (3), (4), and (5) and the other proofs are obvious.

(1) Follows from Definition 5.

(2) If VðHÞ VðKÞ; then we have N j ðVðHÞÞ ¼ fv 2

VðGÞ : N j ðvÞ VðHÞg fv 2 VðGÞ : N j ðvÞ

Since VðH \ KÞ VðHÞ and VðH \ KÞ VðKÞ, then N j ðvÞ VðHÞ and N j ðvÞ VðKÞ. Thus, we have N j ðVðH \ KÞÞ N j ðVðHÞÞ and N j ðVðH \ KÞÞ N j ðVðKÞÞ. Therefore, N j ðVðHÞÞ \ N j ðVðKÞÞ ¼ fv 2 VðGÞ : N j ðvÞ VðHÞg (4) The proof is similar to (3). (5) If v 2 N j ðVðHÞÞ for every v 2 VðHÞ, there exists N j ðvÞ VðHÞ. Then, for every v 2 VðGÞ À ½VðGÞ À VðHÞ, there exists N j ðvÞ such that N j ðvÞ \ ½VðGÞ À VðHÞ ¼ /. So, v 6 2 N j ½VðGÞ À VðHÞ, v 2 VðGÞ À ½N j ðVðGÞ À VðHÞÞ. Therefore,

h Example 3 (Continue for Example 1). Take j ¼ l (and also j 2 fr; 
 [ ; \l [ ; u; i; \u [ ; \i [ g are similar) .

(1) If VðHÞ ¼ fc; dg, then N l ðVðHÞÞ ¼ fb; c; d; eg and N l ðVðHÞÞ ¼ /.

(2) If VðHÞ ¼ fag and VðKÞ ¼ fa; dg, VðHÞ VðKÞ, then N l ðVðHÞÞ ¼ fb; eg, N l ðVðKÞÞ ¼ fb; c; eg.

VðHÞ ¼ fa; bg and VðKÞ ¼ fa; b; dg, VðHÞ VðKÞ, then N l ðVðHÞÞ ¼ /, N l ðVðKÞÞ ¼ fc; eg. (4) If VðHÞ ¼ fbg and VðKÞ ¼ fa; dg, then N l ðVðH \ KÞÞ ¼ /. Hence, N l ðVðHÞÞ ¼ fb; c; eg and N l ðVðKÞÞ ¼ fa; b; c; d; eg. So, P l ðVðHÞÞ\ P l ðVðKÞÞ ¼ fb; c; eg. Also, P l ðVðH [ KÞÞ ¼ fc; eg. Thus, P l ðVðHÞÞ ¼ / and P l ðVðKÞÞ ¼ feg. Therefore, P l ðVðHÞÞ[ P l ðVðKÞÞ ¼ feg.

Remark 2 According to Nicoletti et al. (2001) ; Zafar and Akram (2018) , we can construct new types of rough sets, say, j-rough graph. So, we can also establish new j-approximation graphs which will be denoted by ðVðGÞ;

All properties of Pawlak rough approximation can also be satisfied by the same manner.

In this section, j-adhesion neighborhoods on graphs are introduced. Also, new types of j-lower (respectively, jupper) approximations will be presented and studied. (1) P r ðxÞ ¼ fy 2 VðGÞ : xR ¼ yRg.

(2) P l ðxÞ ¼ fy 2 VðGÞ : Rx ¼ Ryg.

(3) P 
 [ ðxÞ ¼ fy 2 VðGÞ :

To illustrative Definition 7, we introduce Examples 4 and 5.

Example 4 (Continue for Example 1) Take j ¼ fr; l; 
 [ ; \l [ ; u; i; \u [ ; \i [ g, we have P j ðaÞ ¼ fag; P j ðbÞ ¼ fbg; P j ðcÞ ¼ fcg; P r ðdÞ ¼ fdg; P r ðeÞ ¼ feg.

Example 5 Let G be a directed graph as shown in Fig. 2 . Then, the j-adhesion neighborhoods are (i) If j 2 frg, then P j ðaÞ ¼ fa; dg; P j ðbÞ ¼ fbg; P j ðcÞ ¼ fcg; P j ðdÞ ¼ fa; dg. (ii) If j 2 fl; 
 [ ; \i [ g, then P j ðaÞ ¼ fag; P j ðbÞ ¼ fb; cg; P j ðcÞ ¼ fb; cg; P j ðdÞ ¼ fdg.

(iii) If j 2 fig, then P j ðaÞ ¼ fag; P j ðbÞ ¼ fbg; P j ðcÞ ¼ fcg; P j ðdÞ ¼ fdg. (iv) If j 2 fu; \l [ ; \u [ g, then P j ðaÞ ¼ fa; dg; P j ðbÞ ¼ fb; cg; P j ðcÞ ¼ fb; cg; P j ðdÞ ¼ fa; dg.

Definition 8 Let G ¼ ðVðGÞ; EðGÞÞ be a graph and H be a subgraph of G. The second type of lower approximations and upper approximations of H which will be denoted by P j ðVðHÞÞ and P j ðVðHÞÞ, respectively, is defined by P j ðVðHÞÞ ¼ fx 2 VðGÞ : P j ðxÞ VðHÞg; and P j ðVðHÞÞ ¼ VðHÞ S fx 2 VðGÞ : P j ðxÞ \ VðHÞ 6 ¼ /g.

Let G= (V(G), E(G)) be a graph and H be a subgraph of G. The j-boundary, j-positive, j-negative regions and j-accuracy measure of H in terms of j-adhesion Fig. 2 A directed graph G Granular Computing neighborhood which will be denoted by BND P j , POS P j , NEG P j and a Pj , respectively, are defined by (i) BND P j ðVðHÞÞ ¼ P j ðVðHÞÞ À P j ðVðHÞÞ.

(ii) POS P j ðVðHÞÞ ¼ P j ðVðHÞÞ.

(iii) NEG P j ðVðHÞÞ ¼ VðGÞ À P j ðVðHÞÞ.

(iv) a Pj ðVðHÞÞ ¼ jP j ðVðHÞÞj jP j ðVðHÞÞj , jP j ðVðHÞÞj 6 ¼ 0.

Example 6 (Continue from Example 5). Take j ¼ r. If VðHÞ ¼ fa; cg, then we have (i) P r ðVðHÞÞ ¼ fa; c; dg and P r ðVðHÞÞ ¼ fcg.

(ii) BND P r ðVðHÞÞ ¼ fa; dg; POS P r ðVðHÞÞ ¼ fcg; NEG P r ðVðHÞÞ ¼ fbg and a Pj ðVðHÞÞ ¼ 1 3 . For j ¼ l; 
 [ ; \l [ ; u; i; \u [ ; \i [ , we have the results by similarity.

Theorem 2 Let G ¼ ðVðGÞ; EðGÞÞ be a graph and H and K be subgraphs of G. Then, the following properties are held.

(1) P j ð/Þ ¼ /.

(2) P j ðVðGÞÞ ¼ VðGÞ.

(3) P j ðVðHÞÞ VðHÞ.

(4) If VðHÞ VðKÞ, then P j ðVðHÞÞ P j ðVðKÞÞ.

(5) P j ðP j ðVðHÞÞÞ=P j ðVðHÞÞ. (6) P j ðVðHÞ \ VðKÞÞ=P j ðVðHÞÞ \ P j ðVðKÞÞ. (7) P j ðVðHÞÞ [ P j ðVðKÞÞ P j ðVðHÞ [ VðKÞÞ.

(8) P j ðVðHÞÞ ¼ ðP j ðVðHÞÞ c Þ c .

(9) P j ð/Þ ¼ /.

(10) P j ðVðGÞÞ ¼ VðGÞ.

(11) VðHÞ P j ðVðHÞÞ. (12) If VðHÞ VðKÞ, then P j ðVðHÞÞ P j ðVðKÞÞ. (13) P j ðP j ðVðHÞÞÞ ¼ P j ðVðHÞÞ. (14) P j ðVðHÞ [ VðKÞÞ ¼ P j ðVðHÞÞ[ P j ðVðKÞÞ.

(15) P j ðVðHÞ \ VðKÞÞ P j ðVðHÞÞ \ P j ðVðKÞÞ. (16) P j ðVðHÞÞ ¼ ðP j ðVðHÞÞ c Þ c .

Proof It is sufficient to prove properties (1), (2), (3), (4), (5), (6), (7), and (8) and other proofs are obvious.

(1) P j ð/Þ ¼ fv 2 VðGÞ : P j ðvÞ /g ¼ /.

(2) Follows from property (1) and Definition 8.

(3) Follows from Definition 8. (4) If VðHÞ VðKÞ; then we have P j ðVðHÞÞ ¼ fv 2

VðGÞ : P j ðvÞ VðHÞg fv 2 VðGÞ : P j ðvÞ VðKÞg ¼ P j ðVðKÞÞ. (5) Follows from property (3) and Definition 8.

(6) P j ðVðH \ KÞÞ ¼ fv 2 VðGÞ : P j ðvÞ VðH \ KÞg.

Since VðH \ KÞ VðHÞ and VðH \ KÞ VðKÞ, then P j ðvÞ VðHÞ and P j ðvÞ VðKÞ. Thus, by property (4), we have P j ðVðH \ KÞÞ P j (V(H)) and P j ðVðH \ KÞÞ P j ðVðKÞÞ. Therefore, P j ðVðHÞÞ \ P j ðVðKÞÞ ¼ fv 2 VðGÞ : P j ðvÞ VðHÞg v 2 VðGÞ : P j ðvÞ VðKÞg ¼ fv 2 VðGÞ : P j ðvÞ ðVðHÞ \ VðKÞÞg ¼ fv 2 VðGÞ : P j ðvÞ ðVðH \ KÞÞg ¼ P j ðVðH \ KÞÞ ¼ P j ðVðHÞÞ \P j ðVðKÞÞ. (7) The proof is similar to property (6). (8) If v 2 P j (V(H)) for every v 2 VðHÞ, there exists P j ðvÞ VðHÞ.

Then, for every v 2 VðGÞ À ½VðGÞ À VðHÞ, there exists P j ðvÞ such that P j ðvÞ \½VðGÞ À VðHÞ ¼ /. So, v 6 2 P j ½VðGÞ À VðHÞ, v 2 VðGÞ À ½P j ðVðGÞ À VðHÞÞ. Therefore, P j ðVðHÞÞ ¼ VðGÞ À ½P j ðVðGÞ À VðHÞÞ

Example 7 (Continuing from Example 5) Take j ¼ l. Then (i) If VðHÞ ¼ fc; dg, then P l ðVðHÞÞ ¼ fb; c; dg and P l ðVðHÞÞ ¼ fdg. (ii) If VðHÞ ¼ fag and VðKÞ ¼ fa; bg, VðHÞ VðKÞ, then P l ðVðHÞÞ ¼ fag, P l ðVðKÞÞ ¼ fa; b; cg. (iii) If VðHÞ ¼ fa; bg and VðKÞ ¼ fa; b; dg, VðHÞ VðKÞ, then P l ðVðHÞÞ ¼ fag, P l ðVðKÞÞ ¼ fa; dg. (iv) If VðHÞ ¼ fbg and VðKÞ ¼ fa; cg, then P l ðVðH \ KÞÞ ¼ /. Hence, P l ðVðHÞÞ ¼ fb; cg and P l ðVðKÞÞ ¼ fa; b; cg. So, P l ðVðHÞÞ\ P l ðVðKÞÞ ¼ fb; cg. Also, P l ðVðH [ KÞÞ ¼ fa; b; cg. Thus, P l ðVðHÞÞ ¼ / and P l ðVðKÞÞ ¼ fag. Therefore, P l ðVðHÞÞ[ P l ðVðKÞÞ ¼ fag. (1) BND P j ðVðHÞÞ ¼ P j (V(H)) \P j ðVðGÞ À VðHÞÞ.

(2) BND P j ðVðHÞÞ ¼ BND P j ðVðGÞ À VðHÞÞ.

(3) P j ðVðHÞÞ ¼ VðHÞ [ BND P j (V(H)).

(4) P j ðVðHÞÞ ¼ VðHÞ À BND P j (V(H)). (5) BND P j ðVðHÞÞ\ P j ðVðHÞÞ ¼ /.

(6) BND P j ðVðHÞ [ VðKÞÞ BND P j (V(H)) [ BND P j (V(K)). (7) BND P j ðVðHÞ \ VðKÞÞ BND P j ðVðHÞÞ [ BND P j (V(K)). (8) BND P j ðP j ðVðHÞÞÞ BND P j ðVðHÞ.

(9) BND P j ðP j ðVðHÞÞÞ BND P j (V(H). (10) BND P j ðBND P j ðVðHÞÞÞ BND P j (V(H)).

(1) BND P j ðVðHÞÞ ¼ P j ðVðHÞÞ À P j (V(H)) ¼ P j ðVðHÞÞ\ ðP j ðVðHÞÞÞ c ¼ P j ðVðHÞÞ\ P j ðVðGÞ À VðHÞÞ.

(2) BND P j ðVðHÞÞ ¼ P j ðVðHÞÞ \ P j ðVðGÞ À VðHÞÞ ¼ P j ðVðGÞ À ðVðGÞ À VðHÞÞÞ \ P j ðVðGÞ À VðHÞÞ ¼ BND P j ðVðGÞ À VðHÞÞ.

(

). (5) Follows from Definitions 8 and 9. (6) BND P j ðVðHÞ[ VðKÞÞ ¼ P j ðVðHÞ [ VðKÞÞ \P j ðVðGÞ À ðVðHÞ [VðKÞÞÞ ½P j ðVðHÞÞ [P j (V(K))] \½P j ðVðGÞ À VðHÞÞ \P j ðVðGÞ À VðKÞÞ ¼ ½P j ðVðHÞÞ[ P j ðVðKÞÞ \P j ðVðGÞ À VðHÞÞ \½P j ðVðGÞ À VðKÞÞ ¼ ½ðP j (V(H)) \P j ðVðGÞ À VðHÞÞÞ [ðP j (V(K)) \P j ðVðGÞ À VðKÞÞÞ \P j ðVðGÞ À VðHÞÞ ¼ ½ðP j ðVðHÞÞ\ P j ðVðGÞ À VðHÞÞÞ \P j ðVðGÞ À VðKÞÞ[ ½ðP j ðVðKÞÞ\ P j (V(G) ÀVðKÞÞÞ \ P j ðVðGÞ À VðKÞÞ ¼ ½BND P j (V(H)) \P j ðVðGÞ À VðKÞÞ[ ½BND P j (V(K)) \P j (V(G) ÀVðKÞÞ BND P j (V(H)) [ BND P j (V(K)). Therefore, BND P j (V(H) [ V(K)) BND P j (V(H)) [ BND P j (V(K)). (7) BND P j ðVðHÞ\ VðKÞÞ ¼ P j ðVðHÞ\ V(K)) \P j (V(G) ÀðVðHÞ \VðKÞÞÞ ½P j (V(H)) \ P j (V(K))] \½P j ðVðGÞ À VðHÞÞ [ P j ðVðGÞ À VðKÞÞ ¼ ½P j ðVðHÞÞ\ P j ðVðKÞÞ\ P j ðVðGÞ À VðHÞÞ [½P j ðVðHÞÞ\ P j (V(K)) \P j ðVðGÞ À VðKÞÞ ¼ ½BND P j ðVðHÞÞ\ P j ðVðKÞÞ[ ½BND P j (V(K)) \P j (V(H))] BND P j ðVðHÞÞ[ BND P j (V(K)). (8) BND P j ðP j ðVðHÞÞÞ ¼ P j ðP j (V(H))) \P j (V(G) ÀP j ðVðHÞÞÞ ¼ P j ðVðHÞÞ\ P j ðVðGÞ À P j V(H)) P j ðVðHÞÞ\ P j ðVðGÞ À VðHÞÞ ¼ BND P j (V(H)).

Since V(H) P j (V(H)), then ðP j ðVðHÞÞÞ c ðVðHÞÞ c and hence P j ðVðGÞ À P j (V(H))) P j ðVðGÞ À VðHÞÞ. Thus, BND P j ðP j ðVðHÞÞÞ BND P j (V(H). (9) BND P j ðP j ðVðHÞÞÞ ¼ P j ðP j (V(H))) \P j (V(G) ÀP j (V(H))) P j ðVðHÞÞ\ P j ðVðGÞ À P j (V(H))) P j (V(H)) \P j ðVðGÞ À VðHÞÞ ¼ BND P j (V(H)).

Since P j (V(H)) V(H), then P j ðP j (V(H))) P j (V(H)). So, BND P j ðP j (V(H))) BND P j (V(H).

h Remark 3 According to Nicoletti and Zafar results in (Nicoletti et al. 2001; Zafar and Akram 2018) , new types of rough sets ,say, j-rough graphs and new j-approximation graphs, say, ðVðGÞ; P j Þ, 8 j 2 fr; l; 
 [ ; \l [ ; u; i; \u [ ; \i [ g can be constructed. Also, all Pawlak's rough approximation properties can be studied by the same manner.

In algorithm 1, we establish the system of j-adhesion neighborhoods from any graphs in terms of adjacent vertices and their adjacent matrix. Moreover, two vertices are neighbors if they are adjacent.

In this section, we construct some of Pawlak's concepts in terms of graphs.

Definition 10 Let G ¼ ðVðGÞ; EðGÞÞ be a graph and H be a subgraph of G. Then, we have (i) H Pj -definable (H Pj -exact) if P j ðVðHÞÞ ¼ P j (V(H)).

(ii) H Pj -rough if P j (V(H)) 6 ¼ P j (V(H) 

Internally H Pj -undefinable, if P j ðVðHÞÞ ¼ / and P j (V(H)) 6 ¼ VðGÞ, (iii) Externally H Pj -undefinable, if P j (V(H)) 6 ¼ / and P j ðVðHÞÞ ¼ VðGÞ, (iv) Totally H Pj -undefinable, if P j ðVðHÞÞ ¼ / and P j ðVðHÞÞ ¼ VðGÞ.

Definition 12 Let G ¼ ðVðGÞ; EðGÞÞ be a graph and H be a subgraph of G. A membership function 2 j is called a j-strong if x 2 P j (V(H)). It is called a j-weak membership if x 2 P j (V(H)), 8 j 2 fr; l; 
 [ ; \l [ ; u; i; \u [ ; \i [ g. Lemma 1 Let G ¼ ðVðGÞ; EðGÞÞ be a graph and H be a subgraph of G. Then, we have (i) If x2 j VðHÞ, then x 2 VðHÞ. (ii) If x 2 VðHÞ, then x2 j VðHÞ.

Proof Obviously, by Definition 12. h

The inverse implication of Lemma 1 may not be true, in general.

Example 9 (Continue from Example 5) Take j ¼ r. Suppose that VðHÞ ¼ fb; c; dg. Then, P r ðVðHÞÞ ¼ fb; cg and P r ðVðHÞÞ ¼ VðGÞ. It is clear that d 2 VðHÞ, while, 6 2 r V(H) and a2 r V(H) and a 6 2 VðHÞ. We have results for j 2 fl; 
 [ ; \l [ ; u; i; \u [ ; \i [ g by the same manner.

Proposition 2 Let G ¼ ðVðGÞ; EðGÞÞ be a graph and H be a subgraph of G. Then, the following implications are held:

(1) x2 u VðHÞ ) x2 r VðHÞ ) x2 i V(H).

(2) x2 u VðHÞ ) x2 l VðHÞ )x2 i VðHÞ. (5) x2 i VðHÞ ) x2 r VðHÞ )x2 u VðHÞ. (6) x2 i VðHÞ ) x2 l VðHÞ )x2 u VðHÞ.

The converse may not be true, in general.

Example 10 (Continue from Example 5). Suppose that VðHÞ ¼ fa; bg. Then P u ðVðHÞÞ ¼ /, P r ðVðHÞÞ ¼ fbg, P l ðVðHÞÞ ¼ fag and P i ðVðHÞÞ ¼ fa; bg. Accordingly, b 2 2 r ðVðHÞÞ and a 2 2 l (V(H)). But, b 2 6 2 u ðVðHÞÞ and a 2 6 2 u (V(H)). Also, a 2 2 i (V(H)) and b 2 2 i (V(H)). While, a 2 6 2 r (V(H)) and b 2 6 2 l (V(H)).

According to Definitions 8 and 9, we give an algorithm 2 to determine P j ðVðHÞÞ, P j ðVðHÞÞ, BND P j ðVðHÞÞ, POS P j ðVðHÞÞ and NEG P j ðVðHÞÞ.

The comparison in Table 1 between Nada's method (Nada et al. 2018 ) and the proposed method aims to increase the accuracy measure and reduce the boundary region by increasing lower approximations and decreasing the upper approximations. So, Example 11 will be studied at j ¼ r.

Example 11 (Continue for Example 5) Lower approximations and upper approximations in Definitions 5 and 8 are given. Also, the j-boundary and j-accuracy are evaluated the comparison between them are discussed in Table 1 .

In Example 12, we apply Definitions 5, 6, 8 and 9 in Walczak's example in Chemistry. We take five amino acids as a sample in Table 2 . From Table 3 and Fig. 4 , we show that the vertices of subgraphs VðH 1 Þ ¼ fv 1 ; v 4 g and VðH 2 Þ ¼ fv 2 ; v 5 g are necessary to determine the high energy of unfolding. Example 12 (A chemical example) Let VðGÞ ¼ fv 1 ; v 2 ; v 3 ; v 4 ; v 5 g be five amino acids (AAs) which described in terms of five attributes: a 1 ¼ PIE, a 2 ¼ SAC ¼ surface area, a 3 ¼ MR ¼ molecular refractivity, a 4 ¼ LAM ¼ the side chain polarity and a 5 ¼ Vol ¼ molecular volume (Walzak et al. 1999 ) as shown in Table 2 . We illustrate five graphs G k , where k ¼ 1; 2; Á Á Á ; 5 on VðHÞ VðGÞ in Fig. 3 such that R k ¼ fðx i ; x j Þ 2 X Â X : x i ða k Þ À x j ða k Þ \ r k 2 ; i; j; k ¼ 1; 2; Á Á Á ; 5g, where r k represents the standard of the quantitative attributes a k , k ¼ 1; 2; 3; 4; 5. From intersection of five graphs, we have a new graph G in Fig. 4 which uses to construct j-neighborhood systems and j-adhesion neighborhoods for adjacent vertices.

Take j ¼ r. Then, we have (i) N r ðx 1 Þ ¼ fx 1 ; x 4 g; N r ðx 2 Þ ¼ fx 2 ; x 5 g; N r ðx 3 Þ ¼ fx 3 ; x 4 ; x 5 g; N r ðx 4 Þ ¼ fx 4 g; N r ðx 5 Þ ¼ fx 5 g. (ii) P r ðx 1 Þ ¼ fx 1 g; P r ðx 2 Þ ¼ fx 2 g; P r ðx 3 Þ ¼ fx 3 g; P r ðx 4 Þ ¼ fx 4 g; P r ðx 5 Þ ¼ fx 5 g.

Similarily, we obtain the results of j ¼ fl; 
 [ ; \l [ ; u; i; \u [ ; \i [ g. Table 3 gives a comparison between the r-lower approximations, r-upper approximations, r-boundary regions and r-accuracy measures for Nada method at j ¼ r and our proposed method in Definition 8. We prove that Table 3 Comparison between boundary region and accuracy measure for Nada et al's method and our study

Nada's method (Nada et al. 2018 ) (as in Definition 1) (i.e., Definition 5 when j ¼ j r ) 1 3 fv 3 ; v 5 g fv 3 ; v 5 g / 1 Fig. 3 A graph G k for R k , where k ¼ 1; 2; 3; 4; 5

our study method has more accurate than the previous approaches.

7 Conclusion and future work j-Adhesion neighborhoods on general graphs are important tools to approximate graphs as finite structures. Different eight types of j-adhesion neighborhoods are introduced and discussed for each j 2 fr; l; i; u; 
 [ ; \l [ ; \i [ ; \u [ g. By these neighborhoods, j-lower approximations and j-upper approximations will be constructed. Moreover, the relationships among j-approximations are superposed. Furthermore, we show that boundary regions are decreased through increasing the j-lower approximations and decreasing the j-upper approximations. So, the j-accuracy is more accurate than the other type defined in (Nada et al. 2018) . The results in this article are very significant in decision-making, especially, to classify the family of coronavirus (Lai et al. 2020; Kampf et al. 2020) which is a topological space and type of Stone-Cech compactification.

On j-near closure operators induced from relations and its applications

Generalizing covering approximation space

Comparison between different kinds of approximations by using a family of binary relations

Generalization of Pawlak s rough approximation spaces by using cb-open sets

New approach for basic rough set concepts

Multi-criteria decision-making methods under soft rough fuzzy knowledge

On j-near concepts in rough sets with some applications

Comparison of six types of rough approximations based on j-neighborhood space and j-adhesion neighborhood space

Outlier mining in large high-dimensional data sets

Fast feature selection algorithm for neighborhood rough set model based on Bucket and Trie structures

Textbooks in mathematics (graphs and digraphs), 6th edn

Mining constrained gradients in large databases

Reduction based on similarity and decision-making

Some nano topological structures via ideals and graphs

Some topological structures of fractals and their related graphs

On generalization of rough sets by using two different methods

Distance-based information granularity in neighborhood-based granular space

Semantics-preserving dimensionality reduction: rough and fuzzy rough based approaches

Persistence of coronaviruses on inanimate surfaces and their inactivation with biocidal agents

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and coronavirus disease-2019 (COVID-19): the epidemic and the challenges

Knowledge acquisition in incomplete information systems: a rough set approach

Neighborhood attribute reduction approach to partially labeled data

New types of graphs induced by topological spaces

A new approach based on intuitionistic fuzzy rough graphs for decision-making

Soft rough neutrosophic influence graphs with application

Hesitant bipolar-valued fuzzy sets and bipolar-valued hesitant fuzzy sets and their applications in multiattribute group decision making

Intersection of graphs G k to generate a graph G Granular Computing

New types of topological structures via graphs

Rough relation properties

Measurement and indiscernibility

Case generation using rough sets with fuzzy representation

Information systems theoretical foundations

Rough sets

Rough set approach to knowledge-based decision support

Approximation operations in approximation space

Granular computing in decisionmaking

Tolerance approximations spaces

Tutorial rough set theory

Information granulation for rough fuzzy hypergraphs

On generalized rough set theory, rough sets, fuzzy sets, data mining and granular computing

Subsystem based generalizations of rough set approximations

Generalization of rough sets using modal logic

A study of rough sets based on 1-neighborhood systems

A novel decision-making method based on rough fuzzy information

Novel decision-making algorithms based on intuitionistic fuzzy rough environment

Construction of concept granule based on rough set and representation of knowledge-based complex system

Generalized rough sets based on relations

Acknowledgements The authors would like to thank the reviewers for a careful and thorough reading of this manuscript.Funding No applicable.

Conflict of interest No potential conflict of interest was reported by the author.

Our proposed method (as in Definition 8) N r ðVðHÞÞ N r ðVðHÞÞ BND Nr ðVðHÞÞ a Nr ðVðHÞÞ P r ðVðHÞÞ P r ðVðHÞÞ BND Pr ðVðHÞÞ a Pr ðVðHÞÞ