Neutrosophic Sets and Systems, Vol.

17, 2017 3

University of New Mexico

Covering-Based Rough Single Valued Neutrosophic Sets

Yingcang Ma, Wanying Zhou, Qing Wan
School of Science, Xian Polytechnic University, Xian Shaanxi 710048, China. E-mail: mayingcang@126.com, 892443522@qq.com, wqysbe@163.com

Abstract: Rough sets theory is a powerful tool to deal with un-
certainty and incompleteness of knowledge in information systems.
Wang et al. proposed single valued neutrosophic sets as an extension
of intuitionistic fuzzy sets to deal with real-world problems. In this
paper, we propose the covering-based rough single valued neutro-
sophic sets by combining covering-based rough sets and single val-
ued neutrosophic sets. Firstly, three types of covering-based rough
single valued neutrosophic sets models are built and the properties
of lower/upper approximation operators are explored. Secondly, the
lower/upper approximations in two different covering approximation
spaces are studied. The sufficient and necessary condition for gener-
ating the same lower/upper approximations from two different cover-
ing approximation spaces is discussed. Moreover, the relations of the
three models are discussed and the equivalence conditions for three
models are given.

Keywords: covering-based rough sets, single valued neutrosophic sets, neutrosophic sets, covering-based rough single valued neutrosophic sets.

1 Introduction Wei et al. [17]and Xu et al. [18] established the first and sec-
ond types of rough fuzzy set models based on a covering. Hu et
Rough set theory (RST), proposed by Pawlak[1] in 1982, is one al.[19] proposed the third type of rough fuzzy set models based
of the effective mathematical tools for processing fuzzy and un- on a covering. Tang et al. [20] gave the fourth type of rough
certainty knowledge. The classical rough set theory is based on fuzzy set models based on a covering.
the equivalence relation on the domain. In many practical prob- Smarandache [21] proposed neutrosophic sets to deal with
lems, the relation between objects is essentially no equivalence real-world problems. A neutrosophic set has three membership
relation, so this equivalence relation as the basis of the classic functions: truth membership function, indeterminacy member-
rough set model cannot fully meet the actual needs. For this a ship function and falsity membership function, in which each
lot of extension models of Pawlak rough set are given. One ap- membership degree is a real standard or non-standard subset of
proach is to extend the equivalence realtion to similarity relation- the nonstandard unit interval ]0, 1 + [ . Wang et al. [22] intro-
s[2], tolerance relations[3], ordinary binary relations[4], reflex- duced single valued neutrosophic sets (SVNSs) that is a gener-
ive and transitive relations[5] and others. The other approach is alization of intuitionistic fuzzy sets, in which three membership
combining the other theory to get more flexible and expressive functions are independent and their values belong to the unit in-
framework for modeling and processing incomplete information terval [0, 1]. Further studies have done in recent years. Such as,
in information systems. Mi et al.[6] introduced the definitions Majumdar and Samanta [23] studied similarity and entropy of
for generalized fuzzy lower and upper approximation operators SVNSs. Ye [24] proposed correlation coefficients of SVNSs, and
determined by a residual implication. Pei [7] studied generalized applied it to single valued neutrosophic decision-making prob-
fuzzy rough sets. Zhang et al.[8] gave a general framework of in- lems, etc.
tuitionistic fuzzy rough set theory. Yang et al. [9]proposed hesi-
tant fuzzy rough sets and studied the models axiomatic character- SVNSs and covering rough sets are two different tools of deal-
izations by combining hesitant fuzzy sets and rough sets.Zhang et ing with uncertainty information. In order to use the advantages
al.[10] further gave the construction and axiomatic characteriza- of SVNSs and covering rough sets, we establish a hybrid model
tions of interval-valued hesitant fuzzy rough sets, and illustrated of SVNSs and covering rough sets. Broumi and Smarandache
the application of the model. proposed single valued neutrosophic information systems based
Covering rough sets theory is an important rough sets theo- on rough set theory [25]. Yang et al. proposed single valued neu-
ry. Covering rough set model, first proposed by Zakowski[11] trosophic rough set model and single valued neutrosophic refined
in 1983, Bonikowski et al. later studied the structures of cover- rough set model[26,27]. In the present paper, we shall propose
ing[12]. Chen et al. [13]discussed the covering rough sets within covering-based rough single valued neutrosophic sets by fusing
the framework of a completely distributive lattice. Zhu and Wang SVNSs and covering rough sets, and explore a general framework
[14]proposed the reduction of covering rough sets to reduce the of the study of covering-based rough single valued neutrosophic
redundant members in a covering in order to find the small- sets.
est covering. Deng et al. [15] established fuzzy rough set mod- The paper is organized as follows. After this introduction, In
els based on a covering. Li et al. [16] proposed a generalized section 2, we provide the basic notions and operations of Pawlak
fuzzy rough approximation operators based on fuzzy coverings. rough sets, covering rough sets and SVNSs. Based on a SVNR,

Yingcang Ma, Wanying Zhou, Qing Wan. Covering-Based Rough Single Valued Neutrosophic Sets
4 Neutrosophic Sets and Systems, Vol. 17, 2017

Sect. 3 proposes three types of covering-based rough single val- Definition 2.7 Let A and B be two SVNS on U . The union of A
ued neutrosophic sets. Properties of lower/upper approximation and B is a SVNS C, denoted by C = A d B, where u U ,
operators are studied. In Sect. 4, we investigate the relations of TC (u) = max{TA (u), TB (u)}, IC (u) = min{IA (u), IB (u)},
the three types models. The last section summarizes the conclu- FC (u) = min{FA (u), FB (u)}.
sions and gives an outlook for future research. The intersection of A and B is a SVNS D, denoted by D =
A e B, where u U , TD (u) = min{TA (u), TB (u)}, IC (u) =
max{IA (u), IB (u)}, FC (u) = max{FA (u), FB (u)}.
2 Preliminaries
Proposition 2.8 [26] Let A and B be two SVNS on U . The fol-
In this section, we give basic notions and operations on Pawlak lowing results hold:
tough sets, covering-based rough sets and SVNSs. (1) A b A d B and B b A d B;
(2) A e B b A and A e B b B;
Definition 2.1 Let U be a non-empty finite university and R be (3) (Ac )c = A;
an equivalence relations on U . (U, R) is called a Pawlak approx- (4) (A d B)c = Ac e B c ;
imation space. X U , the lower and upper approximations of (5) (A e B)c = Ac d B c .
X, denoted by R(X) and R(X), are defined as follows, respec-
tively:
R(X = {x U |[x]R X}, 3 Covering-based rough neutrosophic
R(X = {x U |[x]R X 6= },
where [x]R = {y U |(x, y) R}. R(X) and R(X) are called sets
as lower and upper approximations operators, respectively. The
Definition 3.1 Let U be a non-empty finite university, C is a cov-
pair (R(X), R(X)) is called a Pawlak rough set.
ering of U , (U, C) be a covering approximation space. A is a
Definition 2.2 Let U be a non-empty finite university, C is a fam- SVNS of U . The first type of lower and upper approximations of
ily of subsets of U . If none subsets in C is empty and C = U , A with respect to (U, C), denoted by F L(A) and F U (A), are
then C is a covering of U . two SVNSs whose membership functions are defined as u U ,
TF L(A) (u) = inf{TA (v)|v M d(u)},
Definition 2.3 Let C be a covering of U , x U . M dC (x) = I F L(A) (u) = sup{IA (v)|v M d(u)},
{K C (S C x S S K K = S)} is called the F F L(A) (u) = sup{FA (v)|v M d(u)},
minimal description of x, When the covering is clear, we omit the T F U (A) (u) = sup{TA (v)|v M d(u)},
lowercase C in the minimal description. I F U (A) (u) = inf{IA (v)|v M d(u)},
FF U (A) (u) = inf{FA (v)|v M d(u)}.
Definition 2.4 Let U be a space of points (objects), with a gener- The pair (F L(A), F U (A)) is called the first type of rough sin-
ic element in U denoted by u. A SVNS A in U is characterized by gle valued neutrosophic set based on covering C. F L(A) and
three membership functions, a truth membership function TA , an F U (A) are called as the first lower and upper approximations
indeterminacy membership function IA and a falsity-membership operators, respectively.
function FA , where u U, TA (u), IA (u), FA (u) [0, 1]. That
is TA : U [0, 1], IA : U [0, 1] and FA : U [0, 1]. Definition 3.2 Let U be a non-empty finite university, C is a cov-
There is no restriction on the sum of TA (u), IA (u) and FA (u), ering of U , (U, C) be a covering approximation space. A is a
thus 0 TA (u) + IA (u) + FA (u) 3. SVNS of U . The second type of lower and upper approximations
of A with respect to (U, C), denoted by SL(A) and SU (A), are
Here A can be denoted by A = two SVNSs whose membership functions are defined as u U ,
{hu, TA (u), IA (u), FA (u)i|u U }, u U, (TA (u), TSL(A) (u) = inf{TA (v)|v M d(u)},
IA (u), FA (u)) is called a single valued neutrosophic num- ISL(A) (u) = sup{IA (v)|v M d(u)},
ber(SVNN). FSL(A) (u) = sup{FA (v)|v M d(u)},
TSU (A) (u) = sup{TA (v)|v M d(u)},
Definition 2.5 Let A and B be two SVNSs on U . If for any u ISU (A) (u) = inf{IA (v)|v M d(u)},
U , TA (u) TB (u), IA (u) IB (u), FA (u) FB (u), then we FSU (A) (u) = inf{FA (v)|v M d(u)}.
called A is contained in B, denoted by A b B. The pair (SL(A), SU (A)) is called the second type of rough s-
ingle valued neutrosophic set based on covering C. SL(A) and
If A b B and B b A, then we called A is equal to B, denoted SU (A) are called as the second lower and upper approximations
by A = B. operators, respectively.
Definition 2.6 Let A be a SVNS on U. The complement of A is Definition 3.3 Let U be a non-empty finite university, C is a cov-
denoted by Ac , where u U , TAc (u) = FA (u), IAc (u) = ering of U , (U, C) be a covering approximation space. A is a
1 IA (u), FAc (u) = TA (u). SVNS of U . The third type of lower and upper approximations

Yingcang Ma, Wanying Zhou, Qing Wan. Covering-Based Rough Single Valued Neutrosophic Sets
Neutrosophic Sets and Systems, Vol. 17, 2017 5

of A with respect to (U, C), denoted by T L(A) and T U (A), are FF L(A) (c) = sup{FA (v)|v M d(c)} = sup{FA (b),
two SVNSs whose membership functions are defined as u U , FA (c), FA (d)} = sup{1, 0, 0.5} = 1.
TT L(A) (u) = supKM d(u) {inf vK {TA (v)}}, FF L(A) (d) = sup{FA (v)|v M d(d)} = sup{FA (c),
IT L(A) (u) = inf KM d(u) {supvK {IA (v)}}, FA (d))} = sup{0, 0.5} = 0.5.
FT L(A) (u) = inf KM d(u) {supvK {FA (v)}}. FF U (A) (a) = inf{FA (v)|v M d(a)} = inf{FA (a),
TT U (A) (u) = inf KM d(u) {supvK {TA (v)}}, FA (b)} = inf{0.1, 1} = 0.1.
IT U (A) (u) = supKM d(u) {inf vK {IA (v)}}, FF U (A) (b) = inf{FA (v)|v M d(b)} = inf{FA (a),
FT U (A) (u) = supKM d(u) {inf vK {FA (v)}}, FA (b), FA (c)} = inf{0.1, 1, 0} = 0.
The pair (T L(A), T U (A)) is called the third type of rough sin- FF U (A) (c) = inf{FA (v)|v M d(c)} = inf{FA (b),
gle valued neutrosophic set based on covering C. T L(A) and FA (c), FA (d)} = inf{1, 0, 0.5} = 0.
T U (A) are called as the third lower and upper approximations FF U (A) (d) = inf{FA (v)|v M d(d)} = inf{FA (c),
operators, respectively. FA (d))} = inf{0, 0.5} = 0.
Thus,
Example 3.4 Let U = {a, b, c, d}, K1 = {a, b}, K2 = F L(A) = {ha, (0.2, 0.8, 1)i, hb, (0.2, 0.8, 1)i, hc, (0.5, 0.7, 1)i,
{b, c}, K3 = {c, d}, C = {K1 , K2 , K3 }. A single val- hd, (0.5, 0.7, 0.5)i},
ued neutrosophic set A = {ha, (0.2, 0.8, 0.1)i, hb, (1, 0.3, 1)i, F U (A) = {ha, (1, 0.3, 0.1)i, hb, (1, 0.3, 0)i, hc, (1, 0.3, 0)i,
hc, (0.5, 0.3, 0)i, hd, (0.6, 0.7, 0.5)i}, then M d(a) = {{a, b}}, hd, (0.6, 0.3, 0)i}.
M d(b) = {{a, b}, {b, c}}, M d(c) = {{b, c}, {c, d}}, M d(d) = Similarly,
{{c, d}}. Thus, SL(A) = {ha, (0.2, 0.8, 1)i, hb, (1, 0.3, 1)i, hc, (0.5, 0.3, 0)i,
TF L(A) (a) = inf{TA (v)|v M d(a)} = inf{TA (a), hd, (0.5, 0.7, 0.5)i},
TA (b)} = inf{0.2, 1} = 0.2. SU (A) = {ha, (1, 0.3, 0.1)i, hb, (1, 0.3, 1)i, hc, (0.5, 0.3, 0)i,
TF L(A) (b) = inf{TA (v)|v M d(b)} = inf{TA (a), hd, (0.6, 0.3, 0)i}.
TA (b), TA (c)} = inf{0.2, 1, 0.5} = 0.2. T L(A) = {ha, (0.2, 0.8, 1)i, hb, (0.5, 0.3, 1)i, hc, (0.5, 0.3, 0.5)i,
TF L(A) (c) = inf{TA (v)|v M d(c)} = inf{TA (b), hd, (0.5, 0.7, 0.5)i},
TA (c), TA (d)} = inf{1, 0.5, 0.6} = 0.5. T U (A) = {ha, (1, 0.3, 0.1)i, hb, (1, 0.3, 0.1)i, hc, (0.6, 0.3, 0)i,
TF L(A) (d) = inf{TA (v)|v M d(d)} = inf{TA (c), hd, (0.6, 0.3, 0)i}.
TA (d))} = inf{0.5, 0.6} = 0.5.
TF U (A) (a) = sup{TA (v)|v M d(a)} = sup{TA (a), Proposition 3.5 The first type of rough single valued neutro-
TA (b)} = sup{0.2, 1} = 1. sophic lower and upper approximation operators defined in Def-
TF U (A) (b) = sup{TA (v)|v M d(b)} = sup{TA (a), inition 3.1 has the following properties: A, B SV N S(U ),
TA (b), TA (c)} = sup{0.2, 1, 0.5} = 1. (1) F L(U ) = U, F U (U ) = U ;
TF U (A) (c) = sup{TA (v)|v M d(c)} = sup{TA (b), (2) F L() = , F U () = ;
TA (c), TA (d)} = sup{1, 0.5, 0.6} = 1. (3) F L(A) b A b F U (A);
TF U (A) (d) = sup{TA (v)|v M d(d)} = sup{TA (c), (4) F L(A e B) = F L(A) e F L(B), F U (A d B) = F U (A) d
TA (d))} = sup{0.5, 0.6} = 0.6. F L(B);
IF L(A) (a) = sup{IA (v)|v M d(a)} = sup{IA (a), (5) A b B F L(A) b F L(B), A b B F U (A) b
IA (b)} = sup{0.8, 0.3} = 0.8. F U (B);
IF L(A) (b) = sup{IA (v)|v M d(b)} = sup{IA (a), (6) F U (AeB) b F U (A)eF U (B), F L(AdB) c F L(A)d
IA (b), TA (c)} = sup{0.8, 0.3, 0.3} = 0.8. F L(B);
IF L(A) (c) = sup{IA (v)|v M d(c)} = sup{IA (b), (7) F L(Ac ) = (F U (A))c , F U (Ac ) = (F L(A))c .
IA (c), IA (d)} = sup{0.3, 0.3, 0.7} = 0.7.
IF L(A) (d) = sup{IA (v)|v M d(d)} = sup{IA (c), Proof: (1) TF L(U ) (u) = inf{TU (v)|v M d(u)} = 1,
IA (d))} = sup{0.3, 0.7} = 0.7. TF U (U ) (u) = sup{TU (v)|v M d(u)} = 1, IF L(U ) (u) =
IF U (A) (a) = inf{IA (v)|v M d(a)} = inf{IA (a), sup{IU (v)|v M d(u)} = 0, IF U (U ) (u) = inf{IU (v)|v
IA (b)} = inf{0.8, 0.3} = 0.3. M d(u)} = 0, FF L(U ) (u) = sup{FU (v)|v M d(u)} = 0,
IF U (A) (b) = inf{IA (v)|v M d(b)} = inf{IA (a), FF U (U ) (u) = inf{FU (v)|v M d(u)} = 0, thus F L(U ) =
IA (b), IA (c)} = inf{0.8, 0.3, 0.3} = 0.3. U, F U (U ) = U .
IF U (A) (c) = inf{IA (v)|v M d(c)} = inf{IA (b), (2) TF L() (u) = inf{T (v)|v M d(u)} = 0, TF U () (u) =
IA (c), IA (d)} = inf{0.3, 0.3, 0.7} = 0.3. sup{T (v)|v M d(u)} = 0, IF L() (u) = sup{I (v)|v
IF U (A) (d) = inf{IA (v)|v M d(d)} = inf{IA (c), M d(u)} = 1, IF U () (u) = inf{I (v)|v M d(u)} = 1,
IA (d))} = inf{0.3, 0.7} = 0.3. FF L() (u) = sup{F (v)|v M d(u)} = 1, FF U () (u) =
FF L(A) (a) = sup{FA (v)|v M d(a)} = sup{FA (a), inf{F (v)|v M d(u)} = 1, thus F L() = , F U () = .
FA (b)} = sup{0.1, 1} = 1. (3) Being u M d(u), so TF L(A) (u) = inf{TA (v)|v
FF L(A) (b) = sup{FA (v)|v M d(b)} = sup{FA (a), M d(u)} TA (u) TF U (A) (u) = sup{TA (v)|v
FA (b), TA (c)} = sup{0.1, 1, 0} = 1. M d(u)} =, IF L(A) (u) = sup{IA (v)|v M d(u)}

Yingcang Ma, Wanying Zhou, Qing Wan. Covering-Based Rough Single Valued Neutrosophic Sets
6 Neutrosophic Sets and Systems, Vol. 17, 2017

IA (u) IF U (A) (u) = inf{IA (v)|v M d(u)} =,
FF L(A) (u) = sup{FA (v)|v M d(u)} FA (u)
FF U (A) (u) = inf{FA (v)|v M d(u)} =, thus, F L(A) b
A b F U (A).
(4) TF L (A e B)(u) = inf{TAeB (v)|v M d(u)} =
inf{min{TA (v), TB (v)}|v M d(u)} = min{inf{TA (v)|v
M d(u)}, inf{TB (v)}|v M d(u)} = min{TF L(A) (u),
TF L(B) (u)}.
IF L (A e B)(u) = sup{IAeB (v)|v M d(u)}
= sup{max{IA (v), IB (v)}|v M d(u)} =
max{sup{IA (v)|v M d(u)}, sup{IB (v)}|v M d(u)}
= max{IF L(A) (u), IF L(B) (u)}.
FF L (A e B)(u) = sup{FAeB (v)|v M d(u)}
= sup{max{FA (v), FB (v)}|v M d(u)} =
max{sup{FA (v)|v M d(u)}, sup{FB (v)}|v M d(u)}
= max{FF L(A) (u), FF L(B) (u)}. Thus, F L(A e B) =
F L(A) e F L(B).
TF U (A d B)(u) = sup{TAdB (v)|v M d(u)}
= sup{max{TA (v), TB (v)}|v M d(u)} =
max{sup{TA (v)|v M d(u)}, sup{TB (v)}|v M d(u)}
= max{TF U (A) (u), TF U (B) (u)}.
IF U (A d B)(u) = inf{IAdB (v)|v M d(u)}
= inf{min{IA (v), IB (v)}|v M d(u)} =
min{inf{IA (v)|v M d(u)}, inf{IB (v)}|v M d(u)}
= min{IF U (A) (u), IF U (B) (u)}.
FF U (A d B)(u) = inf{FAdB (v)|v M d(u)} =
inf{min{FA (v), FB (v)}|v M d(u)} = min{inf{FA (v)|v
M d(u)}, inf{FB (v)}|v M d(u)} = min{FF L(A) (u),
FF L(B) (u)}. Thus, F L(A d B) = F L(A) d F L(B).
So (4) holds.
(5) If A b B, then TF L(A) (u) = inf{TA (v)|v M d(u)}
inf{TB (v)|v M d(u)} = TF L(B) (u), IF L(A) (u) =
sup{IA (v)|v M d(u)} sup{IB (v)|v M d(u)} = let Ki M d(u), TT L(A) (u) = inf vi Ki {TA (vi )}, then
IF L(B) (u), FF L(A) (u) = sup{FA (v)|v M d(u)}
sup{FB (v)|v M d(u)} = FF L(B) (u). So, F L(A) b
F L(B).
The similar method we can get A b B F U (A) b F U (B).
So (5) holds.
(6) Being A e B b A b A d B, A e B b B b A d B, from
(5), (6) holds.
(7) TF L(Ac ) (u) = inf{TAc (v)|v M d(u)} =
inf{FA (v)|v M d(u)} = FF U (A) (u) = T(F U (A))c )(u).
IF L(Ac ) (u) = sup{IAc (v)|v M d(u)} = sup{1
IA (v)|v M d(u)} = 1 inf{IA (v)|v M d(u)} = = inf{supv1 K1 {IA (v1 )},
1 IF U (A) )(u) = I(F U (A))c (u).
FF L(Ac ) (u) = sup{FAc (v)|v M d(u)} = sup{TA (v)|v
M d(u)} = TF U (A) (u) = F(F U (A))c )(u).
So, F L(Ac ) = (F U (A))c . The similar method we can get
F U (Ac ) = (F L(A))c , thus (7) holds.
Remark: F L(F L(A)) = F L(A) and F U (F U (A)) =
F U (A) do not hold generally.
Similarly, we can get the following proposition.
Proposition 3.6 The second type of rough single valued neutro- =
sophic lower and upper approximation operators defined in Def- supvm Km {IT L(A)(vm ) }} = IT L(A) (u).
inition 3.2 has the following properties: A, B SV N S(U ),
(1) SL(U ) = U, SU (U ) = U ;
(2) SL() = , SU () = ;
(3) SL(A) b A b SU (A);
(4) SL(A e B) = SL(A) e SL(B), SU (A d B) = SU (A) d
SL(B);
(5) A b B SL(A) b SL(B), A b B SU (A) b
SU (B);
(6) SU (A e B) b SU (A) e SU (B), SL(A d B) c SL(A) d
SL(B);
(7) SL(Ac ) = (SU (A))c , SU (Ac ) = (SL(A))c .
Proposition 3.7 The third type of rough single valued neutro-
sophic lower and upper approximation operators defined in Def-
inition 3.3 has the following properties: A, B SV N S(U ),
(1) T L(U ) = U, T U (U ) = U ;
(2) T L() = , T U () = ;
(3) T L(A) b A b T U (A);
(4) A b B T L(A) b T L(B), A b B T U (A) b
T U (B);
(5) T U (A e B) b T U (A) e F U (B), T L(A d B) c T L(A) d
T L(B);
(6) T L(Ac ) = (T U (A))c , T U (Ac ) = (T L(A))c .
(7) T L(T L(A)) = T L(A), T U (T U (A)) = T U (A).
Proof: The proofs of (1)-(6) are similar to the Proposition 3.5,
we only show (7).
Let u U, M d(u) = {K1 , K2 , , Km }.
TT L(A) (u) = supKM d(u) {inf vK (T(A) (v))}
= sup{inf v1 K1 {TA (v1 )}, inf v2 K2 {TA (v2 )},
, inf vm Km {TA (vm )}, }. Without loss of generality,
for j 6= i, inf vi Ki {TA (vi )} inf vj Kj {TA (vj )}. Let
vi Ki , from Definition 3.3, we have TT L(A) (vi ) =
supKM d(vi ) {inf vK (T( A)(v))} inf vi Ki (T( A)(vi ))
= TT L(A) (u). Being vi Ki (TT L(A) (vi ) TT L(A)(u) ), so
inf vi Ki {TT L(A)(vi ) } = TT L(A) (u). Let vj Kj , j 6= i,
so inf yj Kj {TT L(A) (vj )} TT L(A) )(u) holds. Thus,
TT L(T L(A)) (u) = supKM d(u) {inf vK {TT L(A) (v)}}
= sup{inf v1 K1 {TT L(A)(v1 ) }, inf v2 K2 {TT L(A)(v2 ) }, ,
inf vm Km {TT L(A)(vm ) }} = TT L(A) (u).
IT L(A) (u) = inf KM d(u) {supvK (I(A) (v))}
supv2 K2 {IA (v2 )},
, supvm Km {IA (vm )}, }. Without loss of generality,
let Ki M d(u), IT L(A) (u) = supvi Ki {IA (vi )}, then
for j 6= i, supvi Ki {IA (vi )} supvj Kj {IA (vj )}. Let
vi Ki , from Definition 3.3, we have IT L(A) (vi ) =
inf KM d(vi ) {supvK (I( A)(v))} supvi Ki (I( A)(vi )) =
IT L(A) (u). Being vi Ki (IT L(A) (vi ) IT L(A)(u) ), so
supvi Ki {IT L(A)(vi ) } = IT L(A) (u). Let vj Kj , j 6= i,
so supyj Kj {IT L(A) (vj )} IT L(A) )(u) holds. Thus,
IT L(T L(A)) (u) = inf KM d(u) {supvK {IT L(A) (v)}}
inf{supv1 K1 {IT L(A)(v1 ) }, supv2 K2 {IT L(A)(v2 ) }, ,

Yingcang Ma, Wanying Zhou, Qing Wan. Covering-Based Rough Single Valued Neutrosophic Sets
Neutrosophic Sets and Systems, Vol. 17, 2017 7

FT L(A) (u) = inf KM d(u) {supvK (F(A) (v))} FT LC1 (A) (u) = inf KM d(u) {sup{FA (v)|v K}},
= inf{supv1 K1 {FA (v1 )}, supv2 K2 {FA (v2 )}, FT LC2 (A) (u) = inf K 0 M d(u) {sup{FA (v)| v K 0 }},
, supvm Km {FA (vm )}, }. Without loss of generality, being C1  C2 , then K 0 M dC2 (u), K
let Ki M d(u), FT L(A) (u) = supvi Ki {FA (vi )}, then M dC1 (u), such that K K 0 , so supvK {FA (v)}
for j 6= i, supvi Ki {FA (vi )} supvj Kj {FA (vj )}. Let supvK 0 {TA (v)}. So inf KM dC1 (u) {supvK {IA (v)}}
vi Ki , from Definition 3.3, we have FT L(A) (vi ) = inf K 0 M dC2 (u) {supvK 0 {FA (v)}}, that is FT LC1 (A)
inf KM d(vi ) {supvK (F( A)(v))} supvi Ki (F( A)(vi )) FT LC2 (A) .
= FT L(A) (u). Being vi Ki (FT L(A) (vi ) FT L(A)(u) ), so Thus we can get T LC2 (A) b T LC1 (A), the similar way we
supvi Ki {FT L(A)(vi ) } = FT L(A) (u). Let vj Kj , j 6= i, can get T UC1 (A) b T UC2 (A). According Proposition 3.7, we
so supyj Kj {FT L(A) (vj )} FT L(A) )(u) holds. Thus, can get T LC2 (A) b T LC1 (A) b A b T UC1 (A) b T UC2 (A)
FT L(T L(A)) (u) = inf KM d(u) {supvK {FT L(A) (v)}} holds.
= inf{supv1 K1 {FT L(A)(v1 ) }, supv2 K2 {FT L(A)(v2 ) } ,
supvm Km {FT L(A)(vm ) }} = FT L(A) (u). Definition 4.3 Let C be a covering of a domain U and K C.
That is, T L(T L(A)) = T L(A), the similar way we can get If K is a union of some sets in C K, we say K is reducible in
T U (T U (A)) = T U (A). So (7) holds. C, otherwise K is irreducible. Let C be a covering of U . If every
Remark: T L(A e B) = T L(A) e T L(B) and T U (A d B) = element in C is irreducible, we say C is irreducible; otherwise C
T U (A) d T L(B) do not hold generally. is reducible. K C, if K is reducible in C, then we can omit
K from C, until C is irreducible, which is called a reduction of
C, denoted by reduct(C).
4 The relations among the three types
of covering-based rough single valued Let (U, C) be a covering approximation space, reduct(C) is
the reduction of C, being u U , M d(u) is same in C and
neutrosophic sets models reduct(C), so C = reduct(C), so we can get the following
result.
Definition 4.1 Let C1 , C2 are two coverings on a non-empty fi-
nite university U , u U , K M dC1 (u), there exists K 0 Proposition 4.4 Let (U, C) be a covering approximation space,
M dC2 (u), such that K 0 K, which is called C2 is thinner than reduct(C) is the reduction of C, then A SV N S(U ), C and
C1 , denoted by C2  C1 . If C2  C1 and C1  C2 , which is reduct(C) generate the same covering-based lower/upper ap-
called C1 equals C2 , denoted by C1 = C2 . otherwise, which is proximations for each type of covering-base rough single valued
called C1 does not equal C2 , denoted by C1 6= C2 . If C2 C1 neutrosophic set.
and C1 6= C2 , it is called C2 is strict thinner than C1 , denoted
by C2 < C1 . If K U, K C1 K C2 , it is called C1 Proposition 4.5 Let C1 , C2 are two coverings on a non-empty
identity to C2 , denoted by C1 C2 . finite university U , then A, the lower/upper approximations for
Proposition 4.2 Let C1 , C2 are two coverings on a non-empty each type of covering-base rough single valued neutrosophic set
finite university U , C1  C2 , A is a single valued neutrosophic are same in (U, C1 ) and (U, C2 ) iff reduct(C1 ) = reduct(C2 ).
set on U . We have:
(1) F LC2 (A) b F LC1 (A) b A b F UC1 (A) b F UC2 (A); Proof: Being reduct(C1 ) = reduct(C2 ), A, A is a single
(2) SLC2 (A) b SLC1 (A) b A b SUC1 (A) b SUC2 (A); valued neutrosophic set on U , from Proposition 4.2 we can get
(3) T LC2 (A) b T LC1 (A) b A b T UC1 (A) b T UC2 (A). the results hold.
We just prove the third types of rough single valued neutro-
Proof: We only show (3). sophic set model, the others are similarly.
Let u U , TT LC1 (A) (u) = supKM d(u) {inf{TA (v)|v Proof by contradiction. Assume reduct(C1 ) 6= reduct(C2 ),
K}}, TT LC2 (A) (u) = supK 0 M d(u) {inf{TA (v)|v K 0 }}, let K reduct(C1 ), K 6 reduct(C2 ). We have
being C1  C2 , then K 0 M dC2 (u), K F Lreduct(C1 ) (K) = K (here K be a single valued neutro-
M dC1 (u), such that K K 0 , so inf vK {TA (v)} sophic set, TK (u) = 1, if u K, otherwise TK (u) = 0.
inf vK 0 {TA (v)}. So supKM dC1 (u) {inf vK {TA (v)}} IK (u) = 0, if u K, otherwise IK (u) = 1. FK (u) = 0,
supK 0 M dC2 (u) {inf vK 0 {TA (v)}}, that is TT LC1 (A) if u K, otherwise FK (u) = 1). From Proposition 4.4, if
TT LC2 (A) . K has the same covering-based rough single valued neutrosoph-
IT LC1 (A) (u) = inf KM d(u) {sup{IA (v)|v K}}, ic set in (U, C1 ) and (U, C2 ), then K has the same covering-
IT LC2 (A) (u) = inf K 0 M d(u) {sup{IA (v)| v K 0 }}, based rough single valued neutrosophic set in (U, reduct(C1 ))
being C1  C2 , then K 0 M dC2 (u), K and (U, reduct(C2 )), so F Lreduct(C2 ) (K) = K. Being K 6
M dC1 (u), such that K K 0 , so supvK {IA (v)} reduct(C2 ), then there exist k1 , k2 , , kn reduct(C2 ), such
supvK 0 {TA (v)}. So inf KM dC1 (u) {supvK {IA (v)}} that K = 1in ki . For each ki reduct(C2 ), there exist
inf K 0 M dC2 (u) {supvK 0 {IA (v)}}, that is IT LC1 (A) ki1 , ki2 , , kimi reduct(C1 ), such that ki = 1jmi kij ,
IT LC2 (A) . so K = 1in 1jmi kij , that is K is reducible in

Yingcang Ma, Wanying Zhou, Qing Wan. Covering-Based Rough Single Valued Neutrosophic Sets
8 Neutrosophic Sets and Systems, Vol. 17, 2017
reduct(C1 ), which is contradiction that reduct(C) is a reduc-
tion of C. So the result holds.
u U, K M d(u), it is obviously that M d(u) K [1] Z. Pawlak. (1982). Rough sets. International Journal of
M d(u), so we can get the following proposition.
Proposition 4.6 Let (U, C) be a covering approximation space,
A is a single valued neutrosophic set, then F L(A) b T L(A) b
SL(A) b A b SU (A) b T U (A) b F U (A).
Proposition 4.7 Let (U, C) be a covering approximation space,
A is a single valued neutrosophic set, then the three types
covering-based rough single valued neutrosophic sets are equiv-
alence iff u U , inf{A(v)|v M d(u)} = inf{A(v)|v
M d(u)} and u U , sup{A(v)|v M d(u)} = [4] Y.Y. Yao. (1998). Relational interpretations of neighbor-
sup{A(v)|v M d(u)}
Proof: From Proposition 4.6 we can get T LC2 (A) b
T LC1 (A) b A b T UC1 (A) b T UC2 (A), being u U ,
inf{A(v)|v M d(u)} = inf{A(v)|v M d(u)}, from Def-
inition 3.1, 3.2, 3.3, we can get F L(A) = SL(A) = T L(A) and
F U (A) = SU (A) = T U (A) .
If the three types covering-based rough single valued neu-
trosophic sets are same, from Definition 3.1, 3.2, 3.3, we can
easily get u U , inf{A(v)|v M d(u)} = inf{A(v)|v
M d(u)} and sup{A(v)|v M d(u)} = sup{A(v)|v
M d(u)}.
5 Conclusion
In this paper, we proposed the hybrid models of single valued
neutrosophic refined sets, covering-based rough sets and
covering-based rough single valued neutrosophic sets.
Specifically, we explored the hybrid models through three
different definitions and give the basic properties. Moreover,
we discussed the relations of the three models. For the future
prospects, we plan to explore the application of the proposed
model to data mining and attribute reduction.
Acknowledgement
This work is partially supported by National Natural Science
Foundation of China (Grant No. 11501435), Scholarship Fund
for Studying Abroad by China Scholarship Council, Scientific
Research Program Funded by Shaanxi Provincial Education De-
partment (Program No.12JK0878) and Science and Technology
Plan Project Funded by Science and Technology Bureau of Xian
City (Program No.CXY1441(2)).
Yingcang Ma, Wanying Zhou, Qing Wan. Covering-Based Rough Single Valued Neutrosophic Sets
References
Computer and Information Science, 11(5), 341-356.
[2] R. Slowinski, and D. Vanderpooten. (2000). A Generalized
Definition of Rough Approximations Based on Similarity.
IEEE Transactions on Knowledge and Data Engineering,
12(2), 331-336.
[3] A. Skowron, and J. Stepaniuk. (1996). Tolerance approx-
imation spaces. Fundamenta Informaticae, 27(2-3), 245-
253.
hood operators and rough set approximation operators. In-
formation Sciences, 111(1-4), 239-259.
[5] K. Qin, J. Yang, and Z. Pei. (2008). Generalized rough sets
based on reflexive and transitive relations. Information Sci-
ences, 178(21), 4138-4141.
[6] J.S. Mi, and W.X. Zhang. (2004). An axiomatic characteri-
zation of a fuzzy generalization of rough sets. Information
Sciences, 160, 235-249.
[7] D.W. Pei. (2005). A generalized model of fuzzy rough sets.
International Journal of General Systems, 34, 603-613.
[8] X.H. Zhang, B. Zhou, and P. Li. (2012). A general frame for
intuitionistic fuzzy rough sets. Information Sciences, 216,
34-49.
[9] X.B. Yang, X.N. Song, Y.S. Qi, and J.Y. Yang. (2014). Con-
structive and axiomatic approaches to hesitant fuzzy rough
set. Soft Computing, 18, 1067C1077.
[10] H.D. Zhang, L. Shu, and S.L. Liao. (2016). On interval-
valued hesitant fuzzy rough approximation operators. Soft
Computing, 20(1), 189C209.
[11] W. Zakowski. (1983). Approximations in the space (U, ).
Demonstration Mathematica, 16, 761C769.
[12] Z. Bonikowski, E. Bryniarski, and U. Wybraniec. (1998).
Extensions and intentions in the rough set theory. Informa-
tion Sciences, 107, 149C167.
[13] D. Chen, W.X. Zhang, D. Yeung, and E.C.C. Tsang. (2006).
Rough approximations on a complete completely distribu-
tive lattice with applications to generalized rough sets. In-
formation Sciences, 176(13), 1829-1848.
[14] W. Zhu, and F.Y. Wang. Reduction and axiomization of cov-
ering generalized rough sets. Information Sciences, 152(1),
217-230.
Neutrosophic Sets and Systems, Vol. 17, 2017
[15] T. Deng, Y. Chen, W. Xu, and Q. Dai. (2007). A novel ap-
proach to fuzzy rough sets based on a fuzzy covering. In-
formation Sciences, 177, 2308C2326.
[16] T.J. Li, Y. Leung, and W.X. Zhang. (2008). Generalized
fuzzy rough approximation operators based on fuzzy cover-
ings. International Journal of Approximate Reasoning, 48,
836C856.
[17] L. Wei, D.Q. Miao, F.F. Xu, and F. Xia. (2006). Research
on a covering rough fuzzy set model. Journal of Computer
Research and Development, 43(10), 1719-1723.
[18] Z.Y. Xu, and J.Q. Liao. (2006). On the covering fuzzy rough
sets model. Fuzzy Systems and Mathematics, 20(3), 141-
144.
[19] J. Hu, G.Y. Wang, and Q.H. Zhang. (2010). Covering based
generalized rough fuzzy set model. Journal of Software,
21(5), 968- 977.
[20] J.G. Tang, K. She, and W. Zhu. (1012). A new type of
covering-based rough fuzzy set model. Control and Deci-
sion, 21(11), 1653-1662.
[21] F. Smarandache. (1998). A Unifying Field in Logics. Neu-
trosophy: Neutrosophic probability, set and logic. Re-
hoboth: American Research Press.
[22] H. Wang, F. Smarandache, Y.Q. Zhang, and R. Sunderra-
man. (2010). Single valued neutrosophic sets. Multispace-
and Multistruct, 4, 410-413.
Yingcang Ma, Wanying Zhou, Qing Wan. Covering-Based Rough Single Valued Neutrosophic Sets
9
[23] P. Majumdar, and S.K. Samant. (2014). On similarity and
entropy of neutrosophic sets. Journal of Intelligent & Fuzzy
Systems, 26(3), 1245-1252.
[24] J. Ye. (2014). Improved correlation coefficients of single
valued neutrosophic sets and interval neutrosophic sets for
multiple attribute decision making. Journal of Intelligent &
Fuzzy Systems, 27, 2453-2462.
[25] A.A. Salama, and S. Broumi. (2014). Roughness of Neutro-
sophic Sets. Elixir Applied Mathematics, 74, 26833-26837.
[26] H.L. Yang, C. L. Zhang, Z.L. Guo, and Y.L. Liu.(2016).
A hybrid model of single valued neutrosophic sets and
rough sets: single valued neutrosophic rough set model.
Soft Computing, doi:10.1007/s00500- 016-2356-y.
[27] Y.L. Bao, and H. L. Yang. (2017). On single valued neutro-
sophic refined rough set model and its application. Journal
of Intelligent & Fuzzy Systems, 33(2), 1235-1248.
[28] Abdel-Basset, M., Mohamed, M., & Sangaiah, A. K. (2017).
Neutrosophic AHP-Delphi Group decision making
model based on trapezoidal neutrosophic numbers.
Journal of Ambient Intelligence and Humanized
Computing,
1-17.https://doi.org/10.1007/s12652-017-0548-7
Received: June 26, 2017. Accepted: July 10, 2017.