Natural language processing (NLP) performs a vital function in text summarization, a task targeted at refining the crucial information from the massive quantity of textual data. NLP methods allow computers to comprehend and process human language, permitting the development of advanced summarization methods. Text summarization includes the automatic generation of a concise and coherent summary of a specified document or collection of documents. Extracting significant insights from text data is crucial as it provides advanced solutions to end-users and business organizations. Automatic text summarization (ATS) computerizes text summarization by decreasing the initial size of the text without the loss of main data features. Deep learning (DL) approaches exhibited significant performance in abstractive and extractive summarization tasks. This research designed an extractive text summarization using NLP with an optimal DL (ETS-NLPODL) model. The major goal of the ETS-NLPODL technique was to exploit feature selection with a hyperparameter-tuned DL model for summarizing the text. In the ETS-NLPODL technique, an initial step of data preprocessing was involved to convert the input text into a compatible format. Next, a feature extraction process was carried out and the optimal set of features was chosen by the hunger games search optimization (HGSO) algorithm. For text summarization, the ETS-NLPODL model used an attention-based convolutional neural network with a gated recurrent unit (ACNN-GRU) model. Finally, the mountain gazelle optimization (MGO) algorithm was employed for the optimal hyperparameter selection of the ACNN-GRU model. The experimental results of the ETS-NLPODL system were examined under the benchmark dataset. The experimentation outcomes pointed out that the ETS-NLPODL technique gained better performance over other methods concerning diverse performance measures.
Citation: Abdulkhaleq Q. A. Hassan, Badriyya B. Al-onazi, Mashael Maashi, Abdulbasit A. Darem, Ibrahim Abunadi, Ahmed Mahmud. Enhancing extractive text summarization using natural language processing with an optimal deep learning model[J]. AIMS Mathematics, 2024, 9(5): 12588-12609. doi: 10.3934/math.2024616
[1] | Gauhar Rahman, Kottakkaran Sooppy Nisar, Feng Qi . Some new inequalities of the Grüss type for conformable fractional integrals. AIMS Mathematics, 2018, 3(4): 575-583. doi: 10.3934/Math.2018.4.575 |
[2] | Muhammad Tariq, Sotiris K. Ntouyas, Hijaz Ahmad, Asif Ali Shaikh, Bandar Almohsen, Evren Hincal . A comprehensive review of Grüss-type fractional integral inequality. AIMS Mathematics, 2024, 9(1): 2244-2281. doi: 10.3934/math.2024112 |
[3] | Da Shi, Ghulam Farid, Abd Elmotaleb A. M. A. Elamin, Wajida Akram, Abdullah A. Alahmari, B. A. Younis . Generalizations of some q-integral inequalities of Hölder, Ostrowski and Grüss type. AIMS Mathematics, 2023, 8(10): 23459-23471. doi: 10.3934/math.20231192 |
[4] | Marwa M. Tharwat, Marwa M. Ahmed, Ammara Nosheen, Khuram Ali Khan, Iram Shahzadi, Dumitru Baleanu, Ahmed A. El-Deeb . Dynamic inequalities of Grüss, Ostrowski and Trapezoid type via diamond-α integrals and Montgomery identity. AIMS Mathematics, 2024, 9(5): 12778-12799. doi: 10.3934/math.2024624 |
[5] | Mustafa Gürbüz, Yakup Taşdan, Erhan Set . Ostrowski type inequalities via the Katugampola fractional integrals. AIMS Mathematics, 2020, 5(1): 42-53. doi: 10.3934/math.2020004 |
[6] | Shuang-Shuang Zhou, Saima Rashid, Asia Rauf, Fahd Jarad, Y. S. Hamed, Khadijah M. Abualnaja . Efficient computations for weighted generalized proportional fractional operators with respect to a monotone function. AIMS Mathematics, 2021, 6(8): 8001-8029. doi: 10.3934/math.2021465 |
[7] | J. Vanterler da C. Sousa, E. Capelas de Oliveira . The Minkowski’s inequality by means of a generalized fractional integral. AIMS Mathematics, 2018, 3(1): 131-147. doi: 10.3934/Math.2018.1.131 |
[8] | Sajid Iqbal, Muhammad Samraiz, Gauhar Rahman, Kottakkaran Sooppy Nisar, Thabet Abdeljawad . Some new Grüss inequalities associated with generalized fractional derivative. AIMS Mathematics, 2023, 8(1): 213-227. doi: 10.3934/math.2023010 |
[9] | Thanin Sitthiwirattham, Muhammad Aamir Ali, Hüseyin Budak, Sotiris K. Ntouyas, Chanon Promsakon . Fractional Ostrowski type inequalities for differentiable harmonically convex functions. AIMS Mathematics, 2022, 7(3): 3939-3958. doi: 10.3934/math.2022217 |
[10] | Ghulam Farid, Hafsa Yasmeen, Hijaz Ahmad, Chahn Yong Jung . Riemann-Liouville Fractional integral operators with respect to increasing functions and strongly (α,m)-convex functions. AIMS Mathematics, 2021, 6(10): 11403-11424. doi: 10.3934/math.2021661 |
Natural language processing (NLP) performs a vital function in text summarization, a task targeted at refining the crucial information from the massive quantity of textual data. NLP methods allow computers to comprehend and process human language, permitting the development of advanced summarization methods. Text summarization includes the automatic generation of a concise and coherent summary of a specified document or collection of documents. Extracting significant insights from text data is crucial as it provides advanced solutions to end-users and business organizations. Automatic text summarization (ATS) computerizes text summarization by decreasing the initial size of the text without the loss of main data features. Deep learning (DL) approaches exhibited significant performance in abstractive and extractive summarization tasks. This research designed an extractive text summarization using NLP with an optimal DL (ETS-NLPODL) model. The major goal of the ETS-NLPODL technique was to exploit feature selection with a hyperparameter-tuned DL model for summarizing the text. In the ETS-NLPODL technique, an initial step of data preprocessing was involved to convert the input text into a compatible format. Next, a feature extraction process was carried out and the optimal set of features was chosen by the hunger games search optimization (HGSO) algorithm. For text summarization, the ETS-NLPODL model used an attention-based convolutional neural network with a gated recurrent unit (ACNN-GRU) model. Finally, the mountain gazelle optimization (MGO) algorithm was employed for the optimal hyperparameter selection of the ACNN-GRU model. The experimental results of the ETS-NLPODL system were examined under the benchmark dataset. The experimentation outcomes pointed out that the ETS-NLPODL technique gained better performance over other methods concerning diverse performance measures.
Hermite and Hadamard's inequality [1,2] is one of the most well-known inequalities in convex function theory, with a geometrical interpretation and numerous applications. The H·H inequality is defined as follows for the convex function Ψ:K→R on an interval K=[u,ν]:
Ψ(u+ν2)≤1ν−u∫νuΨ(x)dx≤Ψ(u)+Ψ(ν)2, | (1) |
for all u,ν∈K.
If f is concave, the inequalities in (1) hold in the reversed direction. We should point out that Hermite-Hadamard inequality is a refinement of the concept of convexity, and it follows naturally from Jensen's inequality. In recent years, the Hermite-Hadamard inequality for convex functions has gotten a lot of attention, and a lot of improvements and generalizations have been examined; see [3,4,5,6,7,8,9,10,11,12] and the references therein.
Interval analysis, on the other hand, is a subset of set-valued analysis, which is the study of sets in the context of mathematical analysis and topology. It was created as a way to deal with interval uncertainty, which can be found in many mathematical or computer models of deterministic real-world phenomena. Archimedes' method, which is used to calculate the circumference of a circle, is a historical example of an interval enclosure. Moore, who is credited with being the first user of intervals in computer mathematics, published the first book on interval analysis in 1966, see [13]. Following the publication of his book, a number of scientists began to research the theory and applications of interval arithmetic. Interval analysis is now a useful technique in a variety of fields that are interested in ambiguous data because of its applications. Computer graphics, experimental and computational physics, error analysis, robotics, and many other fields have applications.
In recent years, several major inequalities (Hermite-Hadamard, Ostrowski, etc.) for interval-valued functions have been studied. Chalco-Cano et al. used the Hukuhara derivative for interval-valued functions to construct Ostrowski type inequalities for interval-valued functions in [14,15]. For interval-valued functions, Román-Flores et al. established Minkowski and Beckenbach's inequalities in [18]. For the rest, see [16,17,18,19,20]. Inequalities, on the other hand, were investigated for the more generic set-valued maps. Sadowska, for example, presented the Hermite-Hadamard inequality in [21]. Other investigations can be found at [22,23].
Recently, Khan et al. [24] introduced the new class of convex fuzzy mappings is known as (h1,h2)-convex F-I-V-Fs by means of FOR such that:
Let h1,h2:[0,1]⊆K=[u,υ]→R+ such that h1,h2≢0. Then F-I-V-F ˜Ψ:K=[u,υ]→FC(R) is said to be (h1,h2)-convex on [u,υ] if
˜Ψ(ξw+(1−ξ)y)≼h1(ξ)h2(1−ξ)˜Ψ(w)˜+h1(1−ξ)h2(ξ)˜Ψ(y), | (2) |
for all w,y∈[u,υ],ξ∈[0,1].
And they also presented the following new version of H·H type inequality for (h1,h2)-convex F-I-V-F involving fuzzy-interval Riemann integrals:
Let ˜Ψ:[u,υ]→F0 be a (h1,h2)-convex F-I-V-F with h1,h2:[0,1]→R+ and h1(12)h2(12)≠0. Then, from θ-levels, we get the collection of I-V-Fs Ψθ:[u,υ]⊂R→K+C are given by Ψθ(ω)=[Ψ∗(ω,θ),Ψ∗(ω,θ)] for all ω∈[u,υ] and for all θ∈[0,1]. If ˜Ψ is fuzzy-interval Riemann integrable (in short, FR-integrable), then
12h1(12)h2(12)˜Ψ(u+υ2)≼1υ−u(FR)∫υu˜Ψ(ω)dω≼[˜Ψ(u)˜+˜Ψ(υ)]∫10h1(ξ)h2(1−ξ)dξ. | (3) |
If h1(ξ)=ξ and h2(ξ)≡1, then from (3), we get following the result for convex F-I-V-F:
˜Ψ(u+υ2)≼1υ−u(FR)∫υu˜Ψ(ω)dω≼˜Ψ(u)˜+˜Ψ(υ)2 |
A one step forward, Khan et al. introduced new classes of convex and generalized convex F-I-V-F, and derived new fractional H·H type and H·H type inequalities for convex F-I-V-F [25], h-convex F-I-V-F [26], (h1,h2)-preinvex F-I-V-F [27], log-s-convex F-I-V-Fs in the second sense [28], LR-log-h-convex I-V-Fs [29], harmonically convex F-I-V-Fs [30], coordinated convex F-I-V-Fs [31] and the references therein. We refer to the readers for further analysis of literature on the applications and properties of fuzzy-interval, and inequalities and generalized convex fuzzy mappings, see [32,33,34,35,36,37,38,39,40,41,42,43,44,45] and the references therein.
The goal of this study is to complete the fuzzy Riemann integrals for interval-valued functions and use these integrals to get the Hermite-Hadamard inequality. These integrals are also used to derive Hermite-Hadamard type inequalities for harmonically convex F-I-V-Fs.
In this section, we recall some basic preliminary notions, definitions and results. With the help of these results, some new basic definitions and results are also discussed.
We begin by recalling the basic notations and definitions. We define interval as,
[ω∗,ω∗]={w∈R:ω∗≤w≤ω∗andω∗,ω∗∈R},whereω∗≤ω∗. |
We write len [ω∗,ω∗]=ω∗−ω∗, If len [ω∗,ω∗]=0 then, [ω∗,ω∗] is called degenerate. In this article, all intervals will be non-degenerate intervals. The collection of all closed and bounded intervals of R is denoted and defined as KC={[ω∗,ω∗]:ω∗,ω∗∈Randω∗≤ω∗}. If ω∗≥0 then, [ω∗,ω∗] is called positive interval. The set of all positive interval is denoted by KC+ and defined as KC+={[ω∗,ω∗]:[ω∗,ω∗]∈KCandω∗≥0}.
We'll now look at some of the properties of intervals using arithmetic operations. Let [ϱ∗,ϱ∗],[s∗,s∗]∈KC and ρ∈R, then we have
[ϱ∗,ϱ∗]+[s∗,s∗]=[ϱ∗+s∗,ϱ∗+s∗], |
[ϱ∗,ϱ∗]×[s∗,s∗]=[min{ϱ∗s∗,ϱ∗s∗,ϱ∗s∗,ϱ∗s∗},max{ϱ∗s∗,ϱ∗s∗,ϱ∗s∗,ϱ∗s∗}], |
ρ.[ϱ∗,ϱ∗]={[ρϱ∗,ρϱ∗]ifρ>0{0}ifρ=0[ρϱ∗,ρϱ∗]ifρ<0. |
For [ϱ∗,ϱ∗],[s∗,s∗]∈KC, the inclusion "⊆" is defined by [ϱ∗,ϱ∗]⊆[s∗,s∗], if and only if s∗≤ϱ∗, ϱ∗≤s∗.
Remark 2.1. The relation "≤I" defined on KC by
[ϱ∗,ϱ∗]≤I[s∗,s∗]if and only ifϱ∗≤s∗,ϱ∗≤s∗, | (4) |
for all [ϱ∗,ϱ∗],[s∗,s∗]∈KC, it is an order relation, see [41]. For given [ϱ∗,ϱ∗],[s∗,s∗]∈KC, we say that [ϱ∗,ϱ∗]≤I[s∗,s∗] if and only if ϱ∗≤s∗,ϱ∗≤s∗ or ϱ∗≤s∗,ϱ∗<s∗.
Moore [13] initially proposed the concept of Riemann integral for I-V-F, which is defined as follows:
Theorem 2.2. [13] If Ψ:[u,ν]⊂R→KC is an I-V-F on such that Ψ(w)=[Ψ∗(w),Ψ∗(w)]. Then Ψ is Riemann integrable over [u,ν] if and only if, Ψ∗ and Ψ∗ both are Riemann integrable over [u,ν] such that
(IR)∫νuΨ(w)dw=[(R)∫νuΨ∗(w)dw,(R)∫νuΨ∗(w)dw]. | (5) |
Let R be the set of real numbers. A mapping ˜ζ:R→[0,1] called the membership function distinguishes a fuzzy subset set A of R. This representation is found to be acceptable in this study. F(R) also stand for the collection of all fuzzy subsets of R.
A real fuzzy interval ˜ζ is a fuzzy set in R with the following properties:
(1) ˜ζ is normal i.e. there exists w∈R such that ˜ζ(w)=1;
(2) ˜ζ is upper semi continuous i.e., for given w∈R, for every w∈R there exist ϵ>0 there exist δ>0 such that ˜ζ(w)−˜ζ(y)<ϵ for all y∈R with |w−y|<δ;
(3) ˜ζ is fuzzy convex i.e., ˜ζ((1−ξ)w+ξy)≥min(˜ζ(w),˜ζ(y)),∀w,y∈R and ξ∈[0,1];
(4) ˜ζ is compactly supported i.e., cl{w∈R|˜ζ(w)>0} is compact.
The collection of all real fuzzy intervals is denoted by F0.
Let ˜ζ∈F0 be real fuzzy interval, if and only if, θ-levels [˜ζ]θ is a nonempty compact convex set of R. This is represented by
[˜ζ]θ={w∈R|˜ζ(w)≥θ}, |
from these definitions, we have
[˜ζ]θ=[ζ∗(θ),ζ∗(θ)], |
where
ζ∗(θ)=inf{w∈R|˜ζ(w)≥θ}, |
ζ∗(θ)=sup{w∈R|˜ζ(w)≥θ}. |
Thus a real fuzzy interval ˜ζ can be identified by a parametrized triples
{(ζ∗(θ),ζ∗(θ),θ):θ∈[0,1]}. |
These two end point functions ζ∗(θ) and ζ∗(θ) are used to characterize a real fuzzy interval as a result.
Proposition 2.3. [18] Let ˜ζ,˜Θ∈F0. Then fuzzy order relation "≼" given on F0 by
˜ζ≼˜Θifandonlyif,[˜ζ]θ≤I[˜Θ]θforallθ∈(0,1], |
it is partial order relation.
We'll now look at some of the properties of fuzzy intervals using arithmetic operations. Let ˜ζ,˜Θ∈F0 and ρ∈R, then we have
[˜ζ˜+˜Θ]θ=[˜ζ]θ+[˜Θ]θ, | (6) |
[˜ζ˜×˜Θ]θ=[˜ζ]θ×[˜Θ]θ, | (7) |
[ρ.˜ζ]θ=ρ.[˜ζ]θ. | (8) |
For ψ∈F0 such that ˜ζ=˜Θ˜+˜ψ, we have the existence of the Hukuhara difference of ˜ζ and ˜Θ, which we call the H-difference of ˜ζ and ˜Θ, and denoted by ˜ζ˜−˜Θ. If H-difference exists, then
(ψ)∗(θ)=(ζ˜−Θ)∗(θ)=ζ∗(θ)−Θ∗(θ),(ψ)∗(θ)=(ζ˜−Θ)∗(θ)=ζ∗(θ)−Θ∗(θ). | (9) |
Definition 2.4. [38] A fuzzy-interval-valued map ˜Ψ:[u,υ]⊂R→F0 is called F-I-V-F. For each θ∈(0,1], whose θ-levels define the family of I-V-Fs Ψθ:[u,υ]⊂R→KC are given by Ψθ(w)=[Ψ∗(w,θ),Ψ∗(w,θ)] for all w∈[u,υ]. Here, for each θ∈(0,1], the end point real functions Ψ∗(.,θ),Ψ∗(.,θ):[u,υ]→R are called lower and upper functions of ˜Ψ.
The following conclusions can be drawn from the preceding literature review [38,39,40]:
Definition 2.5. Let ˜Ψ:[u,ν]⊂R→F0 be an F-I-V-F. Then fuzzy integral of ˜Ψ over [u,ν], denoted by (FR)∫νu˜Ψ(w)dw, it is given level-wise by
[(FR)∫νu˜Ψ(w)dw]θ=(IR)∫νuΨθ(w)dw={∫νuΨ(w,θ)dw:Ψ(w,θ)∈R([u,ν],θ)}, | (10) |
for all θ∈(0,1], where R([u,ν],θ) denotes the collection of Riemannian integrable functions of I-V-Fs. ˜Ψ is FR-integrable over [u,ν] if (FR)∫νu˜Ψ(w)dw∈F0. Note that, if Ψ∗(w,θ),Ψ∗(w,θ) are Lebesgue-integrable, then Ψ is fuzzy Aumann-integrable function over [u,ν], see [18,39,40].
Theorem 2.6. Let ˜Ψ:[u,ν]⊂R→F0 be a F-I-V-F, whose θ-levels define the family of I-V-Fs Ψθ:[u,ν]⊂R→KC are given by Ψθ(w)=[Ψ∗(w,θ),Ψ∗(w,θ)] for all w∈[u,ν] and for all θ∈(0,1]. Then ˜Ψ is FR-integrable over [u,ν] if and only if, Ψ∗(w,θ) and Ψ∗(w,θ) both are R-integrable over [u,ν]. Moreover, if ˜Ψ is FR-integrable over [u,ν], then
[(FR)∫νu˜Ψ(w)dw]θ=[(R)∫νuΨ∗(w,θ)dw,(R)∫νuΨ∗(w,θ)dw]=(IR)∫νuΨθ(w)dw, | (11) |
for all θ∈(0,1]. For all θ∈(0,1],FR([u,ν],θ) denotes the collection of all FR-integrable F-I-V-Fs over [u,ν].
Definition 2.7. [42] A set K=[u,υ]⊂R+=(0,∞) is said to be convex set, if, for all w,y∈K,ξ∈[0,1], we have
wyξw+(1−ξ)y∈K. | (12) |
Definition 2.8. [42] The Ψ:[u,υ]→R+ is called harmonically convex function on [u,υ] if
Ψ(wyξw+(1−ξ)y)≤(1−ξ)Ψ(w)+ξΨ(y), | (13) |
for all w,y∈[u,υ],ξ∈[0,1], where Ψ(w)≥0 for all w∈[u,υ]. If (13) is reversed then, Ψ is called harmonically concave F-I-V-F on [u,υ].
Definition 2.11. [30] The F-I-V-F ˜Ψ:[u,υ]→F0 is called harmonically convex F-I-V-F on [u,υ] if
˜Ψ(wyξw+(1−ξ)y)≼(1−ξ)˜Ψ(w)˜+ξ˜Ψ(y), | (14) |
for all w,y∈[u,υ],ξ∈[0,1], where ˜Ψ(w)≽˜0, for all w∈[u,υ]. If (14) is reversed then, ˜Ψ is called harmonically concave F-I-V-F on [u,υ].
Definition 2.12. The F-I-V-F ˜Ψ:[u,υ]→F0 is called harmonically convex F-I-V-F on [u,υ] if
˜Ψ(wyξw+(1−ξ)y)≼(1−ξ)˜Ψ(w)˜+ξ˜Ψ(y), | (15) |
for all w,y∈[u,υ],ξ∈[0,1], where ˜Ψ(w)≽˜0, for all w∈[u,υ]. If (15) is reversed then, ˜Ψ is called harmonically concave F-I-V-F on [u,υ]. The set of all harmonically convex (harmonically concave) F-I-V-F is denoted by
HFSX([u,υ],F0), |
(HFSV([u,υ],F0)). |
Theorem 2.13. Let [u,υ] be harmonically convex set, and let ˜Ψ:[u,υ]→FC(R) be a F-I-V-F, whose θ-levels define the family of I-V-Fs Ψθ:[u,υ]⊂R→K+C⊂KC are given by
Ψθ(w)=[Ψ∗(w,θ),Ψ∗(w,θ)],∀w∈[u,υ]. | (16) |
for all w∈[u,υ], θ∈[0,1]. Then, ˜Ψ∈HFSX([u,υ],F0), if and only if, for all ∈[0,1],Ψ∗(w,θ), Ψ∗(w,θ)∈HSX([u,υ],R+).
Proof. The demonstration of proof is similar to proof of Theorem 2.12, see [26].
Example 2.14. We consider the F-I-V-Fs ˜Ψ:[0,2]→FC(R) defined by,
˜Ψ(w)(∂)={∂√w∂∈[0,√w]2−∂2√w∂∈(√w,2√w]0otherwise. |
Then, for each θ∈[0,1], we have Ψθ(w)=[θ√w,(2−θ)√w]. Since Ψ∗(w,θ), Ψ∗(w,θ)∈HSX([u,υ],R+), for each θ∈[0,1]. Hence ˜Ψ∈HFSX([u,υ],F0).
Remark 2.15. If T∗(u,θ)=T∗(υ,θ) with θ=1, then harmonically convex F-I-V-F reduces to the classical harmonically convex function, see [42].
In this section, we will prove two types of inequalities. First one is 𝐻.𝐻 and their variant forms, and the second one is H·H Fejér inequalities for convex F-I-V-Fs where the integrands are F-I-V-Fs.
Theorem 3.1. Let ˜Ψ∈HFSX([u,υ],F0), whose θ-levels define the family of I-V-Fs Ψθ:[u,υ]⊂R→K+C are given by Ψθ(w)=[Ψ∗(w,θ),Ψ∗(w,θ)] for all w∈[u,υ], θ∈[0,1]. If ˜Ψ∈FR([u,υ],θ), then
˜Ψ(2uυu+υ)≼uυυ−u∫υu˜Ψ(w)w2dw≼˜Ψ(u)˜+˜Ψ(υ)2. | (17) |
If ˜Ψ∈HFSV([u,υ],F0), then
˜Ψ(2uυu+υ)≽uυυ−u∫υu˜Ψ(w)w2dw≽˜Ψ(u)˜+˜Ψ(υ)2. | (18) |
Proof. Let ˜Ψ∈HFSX([u,υ],F0),. Then, by hypothesis, we have
2˜Ψ(2uυu+υ)≼˜Ψ(uυξu+(1−ξ)υ)˜+˜Ψ(uυ(1−ξ)u+ξυ). |
Therefore, for each θ∈[0,1], we have
2Ψ∗(2uυu+υ,θ)≤Ψ∗(uυξu+(1−ξ)υ,θ)+Ψ∗(uυ(1−ξ)u+ξυ,θ),2Ψ∗(2uυu+υ,θ)≤Ψ∗(uυξu+(1−ξ)υ,θ)+Ψ∗(uυ(1−ξ)u+ξυ,θ). |
Then
2∫10Ψ∗(2uυu+υ,θ)dξ≤∫10Ψ∗(uυξu+(1−ξ)υ,θ)dξ+∫10Ψ∗(uυ(1−ξ)u+ξυ,θ)dξ,2∫10Ψ∗(2uυu+υ,θ)dξ≤∫10Ψ∗(uυξu+(1−ξ)υ,θ)dξ+∫10Ψ∗(uυ(1−ξ)u+ξυ,θ)dξ. |
It follows that
Ψ∗(2uυu+υ,θ)≤uυυ−u∫υuΨ∗(w,θ)w2dw,Ψ∗(2uυu+υ,θ)≤uυυ−u∫υuΨ∗(w,θ)w2dw. |
That is
[Ψ∗(2uυu+υ,θ),Ψ∗(2uυu+υ,θ)]≤Iuυυ−u[∫υuΨ∗(w,θ)w2dw,∫υuΨ∗(w,θ)w2dw]. |
Thus,
˜Ψ(2uυu+υ)≼uυυ−u(FR)∫υu˜Ψ(w)w2dw. | (19) |
In a similar way as above, we have
uυυ−u(FR)∫υu˜Ψ(w)w2dw≼˜Ψ(u)˜+˜Ψ(υ)2. | (20) |
Combining (19) and (20), we have
˜Ψ(2uυu+υ)≼uυυ−u∫υu˜Ψ(w)w2dw≼˜Ψ(u)˜+˜Ψ(υ)2. |
Hence, the required result.
Remark 3.2. If Ψ∗(w,θ)=Ψ∗(w,θ) with θ=1, then Theorem 3.1 reduces to the result for classical harmonically convex function, see [42]:
Ψ(2uυu+υ)≤uυυ−u(R)∫υuΨ(w)w2dw≤Ψ(u)+Ψ(υ)2. |
Example 3.3. We consider the FIVFs ˜Ψ:[0,2]→FC(R), as in Example 2.14. Then, for each θ∈[0,1], we have Ψθ(w)=[θ√w,(2−θ)√w] is harmonically convex FIVF. Since, Ψ∗(w,θ)=θ√w,Ψ∗(w,θ)=(2−θ)√w. We now compute the following:
Ψ∗(2uυu+υ,θ)≤uυυ−u∫υuΨ∗(w,θ)w2dw≤Ψ∗(u,θ)+Ψ∗(υ,θ)2, |
Ψ∗(2uυu+υ,θ)=Ψ∗(0,θ)=0, |
uυυ−u∫υuΨ∗(w,θ)w2dw=02∫20θ√ww2dw=0, |
Ψ∗(u,θ)+Ψ∗(υ,θ)2=θ√2, |
for all θ∈[0,1]. That means
0≤0≤θ√2. |
Similarly, it can be easily show that
Ψ∗(2uυu+υ,θ)≤uυυ−u∫υuΨ∗(w,θ)w2dw≤Ψ∗(u,θ)+Ψ∗(υ,θ)2. |
for all θ∈[0,1], such that
Ψ∗(2uυu+υ,θ)=Ψ∗(0,θ)=0, |
uυυ−u∫υuΨ∗(w,θ)w2dw=02∫20(2−θ)√ww2dw=0, |
Ψ∗(u,θ)+Ψ∗(υ,θ)2=(2−θ)√2. |
From which, we have
0≤0≤(2−θ)√2, |
that is
[0,0]≤I[0,0]≤I1√2[θ,(2−θ)],forallθ∈[0,1]. |
Hence,
˜Ψ(2uυu+υ)≼uυυ−u∫υu˜Ψ(w)w2dw≼˜Ψ(u)˜+˜Ψ(υ)2. |
Theorem 3.4. Let ˜Ψ∈HFSX([u,υ],F0), whose θ-levels define the family of I-V-Fs Ψθ:[u,υ]⊂R→K+C are given by Ψθ(w)=[Ψ∗(w,θ),Ψ∗(w,θ)] for all w∈[u,υ], θ∈[0,1]. If ˜Ψ∈FR([u,υ],θ), then
˜Ψ(2uυu+υ)≼⪧2≼uυυ−u(FR)∫υu˜Ψ(w)w2dw≼⪧1≼˜Ψ(u)˜+˜Ψ(υ)2, | (21) |
where
⪧1=12[˜Ψ(u)˜+˜Ψ(υ)2˜+˜Ψ(2uυu+υ)], |
⪧2=12[˜Ψ(4uυu+3υ)˜+˜Ψ(4uυ3u+υ)], |
and ⪧1=[⪧1∗,⪧1∗], ⪧2=[⪧2∗,⪧2∗]. If ˜Ψ∈HFSV([u,υ],F0), then inequality (21) is reversed.
Proof. Take [u,2uυu+υ], we have
2˜Ψ(u4uυu+υξu+(1−ξ)2uυu+υ+u4uυu+υ(1−ξ)u+ξ2uυu+υ) |
≼˜Ψ(u2uυu+υξu+(1−ξ)2uυu+υ)˜+˜Ψ(u2uυu+υ(1−ξ)u+ξ2uυu+υ). |
Therefore, for every θ∈[0,1], we have
2Ψ∗(u4uυu+υξu+(1−ξ)2uυu+υ+u4uυu+υ(1−ξ)u+ξ2uυu+υ,θ)≤Ψ∗(u2uυu+υξu+(1−ξ)2uυu+υ,θ)+Ψ∗(u2uυu+υ(1−ξ)u+ξ2uυu+υ,θ),2Ψ∗(u4uυu+υξu+(1−ξ)2uυu+υ+u4uυu+υ(1−ξ)u+ξ2uυu+υ,θ)≤Ψ∗(u2uυu+υξu+(1−ξ)2uυu+υ,θ)+Ψ∗(u2uυu+υ(1−ξ)u+ξ2uυu+υ,θ). |
In consequence, we obtain
12Ψ∗(4uυu+3υ,θ)≤uυυ−u∫2uυu+υuΨ∗(w,θ)w2dw,12Ψ∗(4uυu+3υ,θ)≤uυυ−u∫2uυu+υuΨ∗(w,θ)w2dw. |
That is
12[Ψ∗(4uυu+3υ,θ),Ψ∗(4uυu+3υ,θ)]≤Iuυυ−u[∫2uυu+υuΨ∗(w,θ)w2dw,∫2uυu+υuΨ∗(w,θ)w2dw]. |
It follows that
12˜Ψ(4uυu+3υ)≼uυυ−u∫2uυu+υu˜Ψ(w)w2dw. | (22) |
In a similar way as above, we have
12˜Ψ(4uυ3u+υ)≼uυυ−u∫υ2uυu+υ˜Ψ(w)w2dw. | (23) |
Combining (22) and (23), we have
12[˜Ψ(4uυu+3υ)˜+˜Ψ(4uυ3u+υ)]≼uυυ−u∫υu˜Ψ(w)w2dw. | (24) |
Therefore, for every θ∈[0,1], by using Theorem 3.1, we have
Ψ∗(2uυu+υ,θ)≤12[Ψ∗(4uυu+3υ,θ)+Ψ∗(4uυ3u+υ,θ)],Ψ∗(2uυu+υ,θ)≤12[Ψ∗(4uυu+3υ,θ)+Ψ∗(4uυ3u+υ,θ)], |
=⪧2∗,=⪧2∗, |
≤uυυ−u∫υuΨ∗(w,θ)w2dw,≤uυυ−u∫υuΨ∗(w,θ)w2dw, |
≤12[Ψ∗(u,θ)+Ψ∗(υ,θ)2+Ψ∗(2uυu+υ,θ)],≤12[Ψ∗(u,θ)+Ψ∗(υ,θ)2+Ψ∗(2uυu+υ,θ)], |
=⪧1∗,=⪧1∗, |
≤12[Ψ∗(u,θ)+Ψ∗(υ,θ)2+12(Ψ∗(u,θ)+Ψ∗(υ,θ))],≤12[Ψ∗(u,θ)+Ψ∗(υ,θ)2+12(Ψ∗(u,θ)+Ψ∗(υ,θ))], |
=12[Ψ∗(u,θ)+Ψ∗(υ,θ)],=12[Ψ∗(u,θ)+Ψ∗(υ,θ)], |
that is
˜Ψ(2uυu+υ)≼⪧2≼uυυ−u(FR)∫υu˜Ψ(w)w2dw≼⪧1≼12[˜Ψ(u)˜+˜Ψ(υ)]. |
Theorem 3.5. Let ˜Ψ∈HFSX([u,υ],F0) and ˜P∈HFSX([u,υ],F0), whose θ-levels Ψθ,Pθ:[u,υ]⊂R→K+C are defined by Ψθ(w)=[Ψ∗(w,θ),Ψ∗(w,θ)] and Pθ(w)=[P∗(w,θ),P∗(w,θ)] for all w∈[u,υ], θ∈[0,1], respectively. If ˜Ψ˜×˜P∈FR([u,υ],θ), then
uυυ−u(FR)∫υu˜Ψ(w)˜×˜P(w)w2dw≼˜M(u,υ)3˜+˜N(u,υ)6, |
where ˜M(u,υ)=˜Ψ(u)˜×˜P(u)˜+˜Ψ(υ)˜×˜P(υ),˜N(u,υ)=˜Ψ(u)˜×˜P(υ)˜+˜Ψ(υ)˜×˜P(u), and Mθ(u,υ)=[M∗((u,υ),θ),M∗((u,υ),θ)] and Nθ(u,υ)=[N∗((u,υ),θ),N∗((u,υ),θ)].
Proof. Since ˜Ψ,˜P are harmonically convex F-I-V-Fs then, for each θ∈[0,1] we have
Ψ∗(uυ(1−ξ)u+ξυ,θ)≤ξΨ∗(u,θ)+(1−ξ)Ψ∗(υ,θ),Ψ∗(uυ(1−ξ)u+ξυ,θ)≤ξΨ∗(u,θ)+(1−ξ)Ψ∗(υ,θ). |
And
P∗(uυ(1−ξ)u+ξυ,θ)≤ξP∗(u,θ)+(1−ξ)P∗(υ,θ),P∗(uυ(1−ξ)u+ξυ,θ)≤ξP∗(u,θ)+(1−ξ)P∗(υ,θ). |
From the definition of harmonically convexity of F-I-V-Fs it follows that ˜Ψ(w)≽˜0 and ˜P(w)≽˜0, so
Ψ∗(uυ(1−ξ)u+ξυ,θ)×P∗(uυ(1−ξ)u+ξυ,θ)≤(ξΨ∗(u,θ)+(1−ξ)Ψ∗(υ,θ))(ξP∗(u,θ)+(1−ξ)P∗(υ,θ))=Ψ∗(u,θ)×P∗(u,θ)[(ξ)(ξ)]+Ψ∗(υ,θ)×P∗(υ,θ)[(1−ξ)(1−ξ)]+Ψ∗(u,θ)P∗(υ,θ)ξ(1−ξ)+Ψ∗(υ,θ)×P∗(u,θ)ξ(1−ξ), |
Ψ∗(uυ(1−ξ)u+ξυ,θ)×P∗(uυ(1−ξ)u+ξυ,θ)≤(ξΨ∗(u,θ)+(1−ξ)Ψ∗(υ,θ))(ξP∗(u,θ)+(1−ξ)P∗(υ,θ))=Ψ∗(u,θ)×P∗(u,θ)[(ξ)(ξ)]+Ψ∗(υ,θ)×P∗(υ,θ)[(1−ξ)(1−ξ)]+Ψ∗(u,θ)×P∗(υ,θ)ξ(1−ξ)+Ψ∗(υ,θ)×P∗(u,θ)ξ(1−ξ). |
Integrating both sides of above inequality over [0, 1] we get
∫10Ψ∗(uυ(1−ξ)u+ξυ,θ)×P∗(uυ(1−ξ)u+ξυ,θ)=uυυ−u∫υuΨ∗(w,θ)×P∗(w,θ)w2dw≤(Ψ∗(u,θ)×P∗(u,θ)+Ψ∗(υ,θ)×P∗(υ,θ))∫10(ξ)(ξ)dξ+(Ψ∗(u,θ)×P∗(υ,θ)+Ψ∗(υ,θ)×P∗(u,θ))∫10ξ(1−ξ)dξ,∫10Ψ∗(uυ(1−ξ)u+ξυ,θ)×P∗(uυ(1−ξ)u+ξυ,θ)=uυυ−u∫υuΨ∗(w,θ)×P∗(w,θ)w2dw |
≤(Ψ∗(u,θ)×P∗(u,θ)+Ψ∗(υ,θ)×P∗(υ,θ))∫10(ξ)(ξ)dξ |
+(Ψ∗(u,θ)×P∗(υ,θ)+Ψ∗(υ,θ)×P∗(u,θ))∫10ξ(1−ξ)dξ. |
It follows that,
uυυ−u∫υuΨ∗(w,θ)×P∗(w,θ)dw≤M∗((u,υ),θ)3+N∗((u,υ),θ)6uυυ−u∫υuΨ∗(w,θ)×P∗(w,θ)dw≤M∗((u,υ),θ)3+N∗((u,υ),θ)6, |
that is
uυυ−u[∫υuΨ∗(w,θ)×P∗(w,θ)dw,∫υuΨ∗(w,θ)×P∗(w,θ)dw] |
≤I13[M∗((u,υ),θ),M∗((u,υ),θ)]+16[N∗((u,υ),θ),N∗((u,υ),θ)]. |
Thus,
uυυ−u(FR)∫υu˜Ψ(w)˜×˜P(w)w2dw≼˜M(u,υ)3˜+˜N(u,υ)6. |
Theorem 3.6. Let ˜Ψ,˜P∈HFSX([u,υ],F0), whose θ-levels Ψθ,Pθ:[u,υ]⊂R→K+C are defined by Ψθ(w)=[Ψ∗(w,θ),Ψ∗(w,θ)] and Pθ(w)=[P∗(w,θ),P∗(w,θ)] for all w∈[u,υ], θ∈[0,1], respectively. If ˜Ψ˜×˜P∈FR([u,υ],θ), then
2˜Ψ(2uυu+υ)˜×˜P(2uυu+υ)≼uυυ−u(FR)∫υu˜Ψ(w)˜×˜P(w)w2dw+˜M(u,υ)6˜+˜N(u,υ)3, |
where ˜M(u,υ)=˜Ψ(u)˜×˜P(u)˜+˜Ψ(υ)˜×˜P(υ),˜N(u,υ)=˜Ψ(u)˜×˜P(υ)˜+˜Ψ(υ)˜×˜P(u), and Mθ(u,υ)=[M∗((u,υ),θ),M∗((u,υ),θ)] and Nθ(u,υ)=[N∗((u,υ),θ),N∗((u,υ),θ)].
Proof. By hypothesis, for each θ∈[0,1], we have
Ψ∗(2uυu+υ,θ)×J∗(2uυu+υ,θ)Ψ∗(2uυu+υ,θ)×J∗(2uυu+υ,θ) |
≤14[Ψ∗(uυξu+(1−ξ)υ,θ)×J∗(uυξu+(1−ξ)υ,θ)+Ψ∗(uυξu+(1−ξ)υ,θ)×J∗(uυξu+(1−ξ)υ,θ)]+14[Ψ∗(uυ(1−ξ)u+ξυ,θ)×J∗(uυξu+(1−ξ)υ,θ)+Ψ∗(uυξu+(1−ξ)υ,θ)×J∗(uυ(1−ξ)u+ξυ,θ)],≤14[Ψ∗(uυξu+(1−ξ)υ,θ)×J∗(uυξu+(1−ξ)υ,θ)+Ψ∗(uυξu+(1−ξ)υ,θ)×J∗(uυξu+(1−ξ)υ,θ)]+14[Ψ∗(uυ(1−ξ)u+ξυ,θ)×J∗(uυξu+(1−ξ)υ,θ)+Ψ∗(uυξu+(1−ξ)υ,θ)×J∗(uυ(1−ξ)u+ξυ,θ)], |
≤14[Ψ∗(uυξu+(1−ξ)υ,θ)×J∗(uυξu+(1−ξ)υ,θ)+Ψ∗(uυ(1−ξ)u+ξυ,θ)×J∗(uυξu+(1−ξ)υ,θ)]+14[(ξΨ∗(u,θ)+(1−ξ)Ψ∗(υ,θ))×((1−ξ)J∗(u,θ)+ξJ∗(υ,θ))+((1−ξ)Ψ∗(u,θ)+ξΨ∗(υ,θ))×(ξJ∗(u,θ)+(1−ξ)J∗(υ,θ))],≤14[Ψ∗(uυξu+(1−ξ)υ,θ)×J∗(uυξu+(1−ξ)υ,θ)+Ψ∗(uυξu+(1−ξ)υθ)×J∗(uυξu+(1−ξ)υ,θ)]+14[(ξΨ∗(u,θ)+(1−ξ)Ψ∗(υ,θ))×((1−ξ)J∗(u,θ)+ξJ∗(υ,θ))+((1−ξ)Ψ∗(u,θ)+ξΨ∗(υ,θ))×(ξJ∗(u,θ)+(1−ξ)J∗(υ,θ))], |
=14[Ψ∗(uυξu+(1−ξ)υ,θ)×J∗(uυξu+(1−ξ)υ,θ)+Ψ∗(uυξu+(1−ξ)υ,θ)×J∗(uυξu+(1−ξ)υ,θ)]+14[{(ξ)(ξ)+(1−ξ)(1−ξ)}N∗((u,υ),θ)+{ξ(1−ξ)+ξ(1−ξ)}M∗((u,υ),θ)],=14[Ψ∗(uυξu+(1−ξ)υ,θ)×J∗(uυξu+(1−ξ)υ,θ)+Ψ∗(uυξu+(1−ξ)υ,θ)×J∗(uυξu+(1−ξ)υ,θ)]+14[{(ξ)(ξ)+(1−ξ)(1−ξ)}N∗((u,υ),θ)+{ξ(1−ξ)+ξ(1−ξ)}M∗((u,υ),θ)]. |
Integrating over [0,1], we have
2Ψ∗(2uυu+υ,θ)×J∗(2uυu+υ,θ)≤1υ−u(R)∫υuΨ∗(w,θ)×J∗(w,θ)dw+M∗((u,υ),θ)∫10ξ(1−ξ)dξ+N∗((u,υ),θ)∫10(ξ)(ξ)dξ,2Ψ∗(2uυu+υ,θ)×J∗(2uυu+υ,θ)≤1υ−u(R)∫υuΨ∗(w,θ)×J∗(w,θ)dw+M∗((u,υ),θ)∫10ξ(1−ξ)dξ+N∗((u,υ),θ)∫10(ξ)(ξ)dξ, |
that is
2˜Ψ(2uυu+υ)˜×˜P(2uυu+υ)≼uυυ−u(FR)∫υu˜Ψ(w)˜×˜P(w)w2dw+˜M(u,υ)6˜+˜N(u,υ)3. |
The theorem has been proved.
First, we will purpose the following inequality linked with the right part of the classical H−H Fejér inequality for harmonically convex F-I-V-Fs through fuzzy order relation, which is said to be 2nd fuzzy H−H Fejér inequality.
Theorem 3.7. (Second fuzzy H−H Fejér inequality) Let ˜Ψ∈HFSX([u,υ],F0), whose θ-levels define the family of I-V-Fs Ψθ:[u,υ]⊂R→K+C are given by Ψθ(w)=[Ψ∗(w,θ),Ψ∗(w,θ)] for all w∈[u,υ], θ∈[0,1]. If ˜Ψ∈FR([u,υ],θ) and ∇:[u,υ]→R,∇(11u+1υ−1w)=∇(w)≥0, then
(FR)∫νu˜Ψ(w)w2∇(w)dw≼˜Ψ(u)˜+˜Ψ(ν)2∫10∇(w)w2dw. | (25) |
If ˜Ψ∈HFSV([u,υ],F0), then inequality (25) is reversed.
Proof. Let Ψ be a harmonically convex F-I-V-F. Then, for each θ∈[0,1], we have
Ψ∗(uυ(1−ξ)u+ξυ,θ)∇(uυ(1−ξ)u+ξυ) |
≤(ξΨ∗(u,θ)+(1−ξ)Ψ∗(ν,θ))∇(uυ(1−ξ)u+ξυ), |
Ψ∗(uυ(1−ξ)u+ξυ,θ)∇(uυ(1−ξ)u+ξυ) |
≤(ξΨ∗(u,θ)+(1−ξ)Ψ∗(ν,θ))∇(uυ(1−ξ)u+ξυ). | (26) |
And
Ψ∗(uυξu+(1−ξ)υ,θ)∇(uυξu+(1−ξ)υ) |
≤((1−ξ)Ψ∗(u,θ)+ξΨ∗(ν,θ))∇(uυξu+(1−ξ)υ), |
Ψ∗(uυξu+(1−ξ)υ,θ)∇(uυξu+(1−ξ)υ) |
≤((1−ξ)Ψ∗(u,θ)+ξΨ∗(ν,θ))∇(uυξu+(1−ξ)υ). | (27) |
After adding (26) and (27), and integrating over [0,1], we get
∫10Ψ∗(uυ(1−ξ)u+ξυ,θ)∇(uυ(1−ξ)u+ξυ)dξ |
+∫10Ψ∗(uυξu+(1−ξ)υ,θ)∇(uυξu+(1−ξ)υ)dξ≤∫10[Ψ∗(u,θ){ξ∇(uυ(1−ξ)u+ξυ)+(1−ξ)∇(uυξu+(1−ξ)υ)}+Ψ∗(ν,θ){(1−ξ)∇(uυ(1−ξ)u+ξυ)+ξ∇(uυξu+(1−ξ)υ)}]dξ,∫10Ψ∗(uυ(1−ξ)u+ξυ,θ)∇(uυ(1−ξ)u+ξυ)dξ+∫10Ψ∗(uυξu+(1−ξ)υ,θ)∇(uυξu+(1−ξ)υ)dξ≤∫10[Ψ∗(u,θ){ξ∇(uυ(1−ξ)u+ξυ)+(1−ξ)∇(uυξu+(1−ξ)υ)}+Ψ∗(ν,θ){(1−ξ)∇(uυ(1−ξ)u+ξυ)+ξ∇(uυξu+(1−ξ)υ)}]dξ. |
=2Ψ∗(u,θ)∫10ξ∇(uυ(1−ξ)u+ξυ)dξ+2Ψ∗(ν,θ)∫10ξ∇(uυξu+(1−ξ)υ)dξ,=2Ψ∗(u,θ)∫10ξ∇(uυ(1−ξ)u+ξυ)dξ+2Ψ∗(ν,θ)∫10ξ∇(uυξu+(1−ξ)υ)dξ. |
Since ∇ is symmetric, then
=2[Ψ∗(u,θ)+Ψ∗(ν,θ)]∫10ξ∇(uυξu+(1−ξ)υ)dξ, |
=2[Ψ∗(u,θ)+Ψ∗(ν,θ)]∫10ξ∇(uυξu+(1−ξ)υ)dξ. | (28) |
Since
∫10Ψ∗(ξu+(1−ξ)ν,θ)∇(uυ(1−ξ)u+ξυ)dξ=∫10Ψ∗((1−ξ)u+ξν,θ)∇(uυξu+(1−ξ)υ)dξ=uυν−u∫νuΨ∗(w,θ)∇(w)dw∫10Ψ∗((1−ξ)u+ξν,θ)∇(uυξu+(1−ξ)υ)dξ=∫10Ψ∗(ξu+(1−ξ)ν,θ)∇(uυξu+(1−ξ)υ)dξ |
=uυν−u∫νuΨ∗(w,θ)∇(w)dw. | (29) |
From (28) and (29), we have
uυν−u∫νuΨ∗(w,θ)∇(w)dw≤[Ψ∗(u,θ)+Ψ∗(ν,θ)]∫10ξ∇(uυξu+(1−ξ)υ)dξ,uυν−u∫νuΨ∗(w,θ)∇(w)dw≤[Ψ∗(u,θ)+Ψ∗(ν,θ)]∫10ξ∇(uυξu+(1−ξ)υ)dξ, |
that is
[uυν−u∫νuΨ∗(w,θ)∇(w)dw,uυν−u∫νuΨ∗(w,θ)∇(w)dw] |
≤I[Ψ∗(u,θ)+Ψ∗(ν,θ),Ψ∗(u,θ)+Ψ∗(ν,θ)]∫10ξ∇(uυξu+(1−ξ)υ)dξ, |
hence
\frac{{u}\upsilon }{\nu -{u}}\left(FR\right){\int }_{{u}}^{\nu }\frac{\widetilde{\varPsi }\left({w}\right)}{{{w}}^{2}}\nabla \left({w}\right)d{w}\preccurlyeq \left[\widetilde{\varPsi }\left({u}\right)\widetilde{+}\widetilde{\varPsi }\left(\nu \right)\right]{\int }_{0}^{1}\xi \nabla \left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }\right)d\xi , |
this concludes the proof.
Next, we construct first H·H Fejér inequality for harmonically convex F-I-V-F, which generalizes first H-H Fejér inequality for harmonically convex function.
Theorem 3.8. (First fuzzy fractional H-H Fejér inequality) Let \widetilde{\varPsi }\in HFSX\left(\left[{u}, \upsilon \right], {\mathbb{F}}_{0}\right) , whose \theta -levels define the family of I-V-Fs {\varPsi }_{\theta}:\left[{u}, \upsilon \right]\subset \mathbb{R}\to {\mathcal{K}}_{C}^{+} are given by {\varPsi }_{\theta}\left({w}\right) = \left[{\varPsi }_{*}\left({w}, \theta\right), {\varPsi }^{*}\left({w}, \theta\right)\right] for all {w}\in \left[{u}, \upsilon \right] , \theta\in \left[0, 1\right] . If \widetilde{\varPsi }\in {\mathcal{F}\mathcal{R}}_{\left(\left[{u}, \nu \right], \theta\right)} and \nabla :\left[{u}, \upsilon \right]\to \mathbb{R}, \nabla \left(\frac{1}{\frac{1}{{u}}+\frac{1}{\upsilon }-\frac{1}{{w}}}\right) = \nabla \left({w}\right)\ge 0, then
\widetilde{\varPsi }\left(\frac{2{u}\nu }{{u}+\nu }\right){\int }_{{u}}^{\nu }\frac{\widetilde{\varPsi }\left({w}\right)}{{{w}}^{2}}d{w}\preccurlyeq \left(FR\right){\int }_{{u}}^{\nu }\frac{\widetilde{\varPsi }\left({w}\right)}{{{w}}^{2}}\nabla \left({w}\right)d{w}\text{.} | (30) |
If \widetilde{\varPsi }\in HFSV\left(\left[{u}, \upsilon \right], {\mathbb{F}}_{0}\right) , then inequality (30) is reversed.
Proof. Since \varPsi is a harmonically convex, then for \theta\in \left[0, 1\right], we have
{\varPsi }_{*}\left(\frac{2{u}\nu }{{u}+\nu }, \theta\right)\le \frac{1}{2}\left(\begin{array}{c}{\varPsi }_{*}\left(\frac{{u}\upsilon }{\left(1-\xi \right){u}+\xi \upsilon }, \theta\right)+{\varPsi }_{*}\left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }, \theta\right)\end{array}\right) |
{\varPsi }^{*}\left(\frac{2{u}\nu }{{u}+\nu }, \theta\right)\le \frac{1}{2}\left(\begin{array}{c}{\varPsi }^{*}\left(\frac{{u}\upsilon }{\left(1-\xi \right){u}+\xi \upsilon }, \theta\right)+{\varPsi }^{*}\left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }, \theta\right)\end{array}\right). | (31) |
By multiplying (31) by \nabla \left(\frac{{u}\upsilon }{\left(1-\xi \right){u}+\xi \upsilon }\right) = \nabla \left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }\right) and integrate it by \xi over \left[0, 1\right], we obtain
{\varPsi }_{*}\left(\frac{2{u}\nu }{{u}+\nu }, \theta\right){\int }_{0}^{1}\nabla \left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }\right)d\xi |
\le \frac{1}{2}\left(\begin{array}{c}{\int }_{0}^{1}{\varPsi }_{*}\left(\frac{{u}\upsilon }{\left(1-\xi \right){u}+\xi \upsilon }, \theta\right)\nabla \left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }\right)d\xi \\ +{\int }_{0}^{1}{\varPsi }_{*}\left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }, \theta\right)\nabla \left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }\right)d\xi \end{array}\right) |
{\varPsi }^{*}\left(\frac{2{u}\nu }{{u}+\nu }, \theta\right){\int }_{0}^{1}\nabla \left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }\right)d\xi |
\le \frac{1}{2}\left(\begin{array}{c}{\int }_{0}^{1}{\varPsi }^{*}\left(\frac{{u}\upsilon }{\left(1-\xi \right){u}+\xi \upsilon }, \theta\right)\nabla \left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }\right)d\xi \\ +{\int }_{0}^{1}{\varPsi }^{*}\left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }, \theta\right)\nabla \left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }\right)d\xi \end{array}\right). | (32) |
Since
{\int }_{0}^{1}{\varPsi }_{*}\left(\frac{{u}\upsilon }{\left(1-\xi \right){u}+\xi \upsilon }, \theta\right)\nabla \left(\frac{{u}\upsilon }{\left(1-\xi \right){u}+\xi \upsilon }\right)d\xi \\ = {\int }_{0}^{1}{\varPsi }_{*}\left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }, \theta\right)\nabla \left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }\right)d\xi \\ = \frac{{u}\upsilon }{\nu -{u}}{\int }_{{u}}^{\nu }{\varPsi }_{*}\left({w}, \theta\right)\nabla \left({w}\right)d{w}, \\{\int }_{0}^{1}{\varPsi }^{*}\left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }, \theta\right)\nabla \left(\frac{{u}\upsilon }{\xi {u}+\left(1-\xi \right)\upsilon }\right)d\xi \\ = {\int }_{0}^{1}{\varPsi }^{*}\left(\frac{{u}\upsilon }{\left(1-\xi \right){u}+\xi \upsilon }, \theta\right)\nabla \left(\frac{{u}\upsilon }{\left(1-\xi \right){u}+\xi \upsilon }\right)d\xi |
= \frac{{u}\upsilon }{\nu -{u}}{\int }_{{u}}^{\nu }{\varPsi }^{*}\left({w}, \theta\right)\nabla \left({w}\right)d{w}. | (33) |
From (32) and (33), we have
{\varPsi }_{*}\left(\frac{2{u}\nu }{{u}+\nu }, \theta\right)\le \frac{1}{{\int }_{{u}}^{\nu }\nabla \left({w}\right)d{w}}{\int }_{{u}}^{\nu }{\varPsi }_{*}\left({w}, \theta\right)\nabla \left({w}\right)d{w}, \\{\varPsi }^{*}\left(\frac{2{u}\nu }{{u}+\nu }, \theta\right)\le \frac{1}{{\int }_{{u}}^{\nu }\nabla \left({w}\right)d{w}}{\int }_{{u}}^{\nu }{\varPsi }^{*}\left({w}, \theta\right)\nabla \left({w}\right)d{w}. |
From which, we have
\left[{\varPsi }_{*}\left(\frac{2{u}\nu }{{u}+\nu }, \theta\right), {\varPsi }^{*}\left(\frac{2{u}\nu }{{u}+\nu }, \theta\right)\right] \\ \le _{I}\frac{1}{{\int }_{{u}}^{\nu }\nabla \left({w}\right)d{w}}\left[{\int }_{{u}}^{\nu }{\varPsi }_{*}\left({w}, \theta\right)\nabla \left({w}\right)d{w}, {\int }_{{u}}^{\nu }{\varPsi }^{*}\left({w}, \theta\right)\nabla \left({w}\right)d{w}\right], |
that is
\widetilde{\varPsi }\left(\frac{2{u}\nu }{{u}+\nu }\right){\int }_{{u}}^{\nu }\frac{\widetilde{\varPsi }\left({w}\right)}{{{w}}^{2}}d{w}\preccurlyeq \left(FR\right){\int }_{{u}}^{\nu }\frac{\widetilde{\varPsi }\left({w}\right)}{{{w}}^{2}}\nabla \left({w}\right)d{w}\text{.} |
Then we complete the proof.
Remark 3.9. If \nabla \left({w}\right) = 1 , then from Theorems 3.7 and 3.8, we obtain inequality (17). If {\varPsi }_{*}\left({w}, \theta\right) = {\varPsi }^{*}\left({w}, \theta\right) with \theta = 1, then Theorems 3.7 and 3.8 reduce to classical first and second classical H-H Fejér inequality for classical harmonically convex function.
Several novel conclusions in convex analysis and associated optimization theory can be obtained using this new class of functions known as harmonically convex F-I-V. The main findings include some new bounds with error estimations via fuzzy Riemann integrals. All of these papers aim to provide new estimations and optimal approaches. But, the main motivation of this paper is that we obtained new method by using fuzzy integrals for harmonically convex F-I-V-Fs calculus. The authors anticipate that this study may inspire more research in a variety of pure and applied sciences fields.
The authors would like to thank the Rector, COMSATS University Islamabad, Islamabad, Pakistan, for providing excellent research and academic environments and this work was supported by Taif University Researches Supporting Project number (TURSP-2020/326), Taif University, Taif, Saudi Arabia, and the authors T. Abdeljawad and B. Abdalla would like to thank Prince Sultan University for APC and for the support through the TAS research lab.
All authors contributed equally to the writing of this paper. All authors read and approved the final manuscript.
The authors declare that they have no competing interests.
[1] | M. Yadav, R. Katarya, A Systematic Survey of Automatic Text Summarization Using Deep Learning Techniques, In Modern Electronics Devices and Communication Systems: Select Proceedings of MEDCOM 2021, 397–405. Singapore: Springer Nature Singapore, 2023. https://doi.org/10.1007/978-981-19-6383-4_31 |
[2] |
Y. M. Wazery, M. E. Saleh, A. Alharbi, A. A. Ali, Abstractive Arabic text summarization based on deep learning, Comput. Intel. Neurosci., 2022. https://doi.org/10.1155/2022/1566890 doi: 10.1155/2022/1566890
![]() |
[3] |
P. J. Goutom, N. Baruah, P. Sonowal, An abstractive text summarization using deep learning in Assamese, Int. J. Inf. Technol., 2023, 1–8. https://doi.org/10.1007/s41870-023-01279-7 doi: 10.1007/s41870-023-01279-7
![]() |
[4] | V. L. Sireesha, Text Summarization for Resource-Poor Languages: Datasets and Models for Multiple Indian Languages (Doctoral dissertation, International Institute of Information Technology Hyderabad), 2023. |
[5] | S. Dhankhar, M. K. Gupta, Automatic Extractive Summarization for English Text: A Brief Survey. In Proceedings of Second Doctoral Symposium on Computational Intelligence: DoSCI 2021, 183–198. Singapore: Springer Singapore, 2021. https://doi.org/10.1007/978-981-16-3346-1_15 |
[6] | B. Shukla, S. Gupta, A. K. Yadav, D. Yadav, Text summarization of legal documents using reinforcement learning: A study, In Intelligent Sustainable Systems: Proceedings of ICISS 2022, 403–414. Singapore: Springer Nature Singapore, 2022. https://doi.org/10.1007/978-981-19-2894-9_30 |
[7] |
B. Baykara, T. Güngör, Turkish abstractive text summarization using pretrained sequence-to-sequence models, Nat. Lang. Eng., 29 (2023), 1275–1304. https://doi.org/10.1017/S1351324922000195 doi: 10.1017/S1351324922000195
![]() |
[8] |
M. Bani-Almarjeh, M. B. Kurdy, Arabic abstractive text summarization using RNN-based and transformer-based architectures, Inform. Process. Manag., 60 (2023), 103227. https://doi.org/10.1016/j.ipm.2022.103227 doi: 10.1016/j.ipm.2022.103227
![]() |
[9] | H. Aliakbarpour, M. T. Manzuri, A. M. Rahmani, Improving the readability and saliency of abstractive text summarization using a combination of deep neural networks equipped with auxiliary attention mechanisms, J. Supercomput., 2022, 1–28. |
[10] |
S. N. Turky, A. S. A. Al-Jumaili, R. K. Hasoun, Deep learning based on different methods for text summary: A survey, J. Al-Qadisiyah Comput. Sci. Math., 13 (2021), 26. https://doi.org/10.29304/jqcm.2021.13.1.766 doi: 10.29304/jqcm.2021.13.1.766
![]() |
[11] |
G. A. Babu, S. Badugu, Deep learning based sequence to sequence model for abstractive Telugu text summarization, Multimed. Tools Appl., 82 (2023), 17075–17096. https://doi.org/10.1007/s11042-022-14099-x doi: 10.1007/s11042-022-14099-x
![]() |
[12] |
S. A. Tripathy, A. Sharmila, Abstractive method-based text summarization using bidirectional long short-term memory and pointer generator mode, J. Appl. Res. Technol., 21 (2023), 73–86. https://doi.org/10.22201/icat.24486736e.2023.21.1.1446 doi: 10.22201/icat.24486736e.2023.21.1.1446
![]() |
[13] |
N. Shafiq, I. Hamid, M. Asif, Q. Nawaz, H. Aljuaid, H. Ali, Abstractive text summarization of low-resourced languages using deep learning, PeerJ Comput. Sci., 9 (2023), e1176. https://doi.org/10.7717/peerj-cs.1176 doi: 10.7717/peerj-cs.1176
![]() |
[14] | R. Karmakar, K. Nirantar, P. Kurunkar, P. Hiremath, D. Chaudhari, Indian regional language abstractive text summarization using attention-based LSTM neural network, In 2021 International Conference on Intelligent Technologies (CONIT), 1–8, IEEE, 2021. https://doi.org/10.1109/CONIT51480.2021.9498309 |
[15] |
R. Rani, D. K. Lobiyal, Document vector embedding based extractive text summarization system for Hindi and English text, Appl. Intell., 2022, 1–20. https://doi.org/10.1007/s10489-021-02871-9 doi: 10.1007/s10489-021-02871-9
![]() |
[16] |
W. Etaiwi, A. Awajan, SemG-TS: Abstractive Arabic text summarization using semantic graph embedding, Mathematics, 10 (2022), 3225. https://doi.org/10.3390/math10183225 doi: 10.3390/math10183225
![]() |
[17] |
R. T. AlTimimi, F. H. AlRubbiay, Multilingual text summarization using deep learning, Int. J. Eng. Adv. Technol., 7 (2021), 29–39. https://doi.org/10.31695/IJERAT.2021.3712 doi: 10.31695/IJERAT.2021.3712
![]() |
[18] | S. V. Moravvej, A. Mirzaei, M. Safayani, Biomedical text summarization using conditional generative adversarial network (CGAN), arXiv preprint arXiv: 2110.11870, 2021. |
[19] |
A. Al Abdulwahid, Software solution for text summarisation using machine learning based Bidirectional Encoder Representations from Transformers algorithm, IET SOFTWARE, 2023. https://doi.org/10.1049/sfw2.12098 doi: 10.1049/sfw2.12098
![]() |
[20] |
B. Muthu, S. Cb, P. M. Kumar, S. N. Kadry, C. H. Hsu, O. Sanjuan, et al., A framework for extractive text summarization based on deep learning modified neural network classifier, ACM T. Asian Low-Reso., 20 (2021), 1–20. https://doi.org/10.1145/3392048 doi: 10.1145/3392048
![]() |
[21] |
D. Izci, S. Ekinci, E. Eker, M. Kayri, Augmented hunger games search algorithm using logarithmic spiral opposition-based learning for function optimization and controller design, J. King Saud University-Eng. Sci., 2022. https://doi.org/10.1016/j.jksues.2022.03.001 doi: 10.1016/j.jksues.2022.03.001
![]() |
[22] |
M. Mafarja, T. Thaher, M. A. Al-Betar, J. Too, M. A. Awadallah, I. Abu Doush, et al., Classification framework for faulty-software using enhanced exploratory whale optimizer-based feature selection scheme and random forest ensemble learning, Appl. Intell., 2023, 1–43. https://doi.org/10.1007/s10489-022-04427-x doi: 10.1007/s10489-022-04427-x
![]() |
[23] |
B. Liu, J. Xu, W. Xia, State-of-Health Estimation for Lithium-Ion Battery Based on an Attention-Based CNN-GRU Model with Reconstructed Feature Series, Int. J. Energy Res., 2023. https://doi.org/10.1155/2023/8569161 doi: 10.1155/2023/8569161
![]() |
[24] |
P. Sarangi, P. Mohapatra, Evolved opposition-based Mountain Gazelle Optimizer to solve optimization problems, J. King Saud Univ-Com., 2023, 101812. https://doi.org/10.1016/j.jksuci.2023.101812 doi: 10.1016/j.jksuci.2023.101812
![]() |
[25] |
H. J. Alshahrani, K. Tarmissi, A. Yafoz, A. Mohamed, M. A. Hamza, I. Yaseen, et al., Applied linguistics with mixed leader optimizer based English text summarization model, Intell. Autom. Soft Co., 36 (2023). https://doi.org/10.32604/iasc.2023.034848 doi: 10.32604/iasc.2023.034848
![]() |
1. | Deepak PACHPATTE, Tariq A. ALJAAİDİ, NEW GENERALIZATION OF REVERSE MINKOWSKI’S INEQUALITY FOR FRACTIONAL INTEGRAL, 2020, 2587-2648, 10.31197/atnaa.756605 | |
2. | Tariq A. Aljaaidi, Deepak B. Pachpatte, The Minkowski’s inequalities via \psi-Riemann–Liouville fractional integral operators, 2020, 0009-725X, 10.1007/s12215-020-00539-w | |
3. | Tariq A. Aljaaidi, Deepak B. Pachpatte, Mohammed S. Abdo, Thongchai Botmart, Hijaz Ahmad, Mohammed A. Almalahi, Saleh S. Redhwan, (k, ψ)-Proportional Fractional Integral Pólya–Szegö- and Grüss-Type Inequalities, 2021, 5, 2504-3110, 172, 10.3390/fractalfract5040172 | |
4. | Wengui Yang, Certain New Chebyshev and Grüss-Type Inequalities for Unified Fractional Integral Operators via an Extended Generalized Mittag-Leffler Function, 2022, 6, 2504-3110, 182, 10.3390/fractalfract6040182 | |
5. | Omar Mutab Alsalami, Soubhagya Kumar Sahoo, Muhammad Tariq, Asif Ali Shaikh, Clemente Cesarano, Kamsing Nonlaopon, Some New Fractional Integral Inequalities Pertaining to Generalized Fractional Integral Operator, 2022, 14, 2073-8994, 1691, 10.3390/sym14081691 | |
6. | Asha B. Nale, Satish K. Panchal, Vaijanath L. Chinchane, Grüss-type fractional inequality via Caputo-Fabrizio integral operator, 2022, 14, 2066-7752, 262, 10.2478/ausm-2022-0018 | |
7. | Tariq A. Aljaaidi, Deepak B. Pachpatte, Wasfi Shatanawi, Mohammed S. Abdo, Kamaleldin Abodayeh, Generalized proportional fractional integral functional bounds in Minkowski’s inequalities, 2021, 2021, 1687-1847, 10.1186/s13662-021-03582-8 | |
8. | Muhammad Tariq, Sotiris K. Ntouyas, Hijaz Ahmad, Asif Ali Shaikh, Bandar Almohsen, Evren Hincal, A comprehensive review of Grüss-type fractional integral inequality, 2023, 9, 2473-6988, 2244, 10.3934/math.2024112 | |
9. | Saleh S. Redhwan, Tariq A. Aljaaidi, Ali Hasan Ali, Maryam Ahmed Alyami, Mona Alsulami, Najla Alghamdi, New Grüss’s inequalities estimates considering the φ-fractional integrals, 2024, 11, 26668181, 100836, 10.1016/j.padiff.2024.100836 | |
10. | Nale Asha B., Satish K. Panchal, L. Chinchane Vaijanath , Certain fractional integral inequalities using generalized Katugampola fractional integral operator, 2020, 8, 23193786, 809, 10.26637/MJM0803/0013 | |
11. | Bhagwat R. Yewale, Deepak B. Pachpatte, 2023, 9781119879671, 45, 10.1002/9781119879831.ch3 | |
12. | Bouharket Benaissa, Noureddine Azzouz, Mehmet Sarikaya, On some Grüss-type inequalities via k-weighted fractional operators, 2024, 38, 0354-5180, 4009, 10.2298/FIL2412009B | |
13. | Xiaohong Zuo, Wengui Yang, Mohammad W. Alomari, Certain Novel p,q‐Fractional Integral Inequalities of Grüss and Chebyshev‐Type on Finite Intervals, 2025, 2025, 2314-4629, 10.1155/jom/9536854 |