The worst-case scenario: robust portfolio optimization with discrete distributions and transaction costs

Ebenezer Fiifi Emire Atta Mills; Ebenezer Fiifi Emire Atta Mills

doi:10.3934/math.20241018

AIMS Mathematics

2024, Volume 9, Issue 8: 20919-20938. doi: 10.3934/math.20241018

Previous Article Next Article

Research article

The worst-case scenario: robust portfolio optimization with discrete distributions and transaction costs

Ebenezer Fiifi Emire Atta Mills ^{1,2
,
,}

1.
School of Mathematical Sciences, Wenzhou-Kean University, Wenzhou, China
2.
Academy of Research for Sustainability (AIRs), Wenzhou-Kean University, Wenzhou, China

Received: 14 March 2024 Revised: 27 May 2024 Accepted: 30 May 2024 Published: 28 June 2024
MSC : 91B05, 91G10

This research introduces min-max portfolio optimization models that incorporating transaction costs and focus on robust Entropic value-at-risk. This study offers a unified approach to handl the distribution of random parameters that affect the reward and risk aspects. Utilizing the duality theorem, the study transforms the optimization models into manageable forms, thereby accommodating the underlying random variables' discrete box and ellipsoidal distributions. The impact of transaction costs on optimal portfolio selection is examined through numerical examples under a robust return-risk framework. The results underscore the importance of the proposed model in safeguarding capital and reducing exposure to extreme risks, thus outperforming other strategies documented in the literature. This demonstrates the model's effectiveness in balancing maximizing returns and minimizing potential losses, making it a valuable tool for investors that seek to navigate uncertain financial markets.

Keywords:

Citation: Ebenezer Fiifi Emire Atta Mills. The worst-case scenario: robust portfolio optimization with discrete distributions and transaction costs[J]. AIMS Mathematics, 2024, 9(8): 20919-20938. doi: 10.3934/math.20241018

Related Papers:

[1]	Ibtesam Alshammari, Islam M. Taha . On fuzzy soft $\beta$ -continuity and $\beta$ -irresoluteness: some new results. AIMS Mathematics, 2024, 9(5): 11304-11319. doi: 10.3934/math.2024554
[2]	R. Mareay, Radwan Abu-Gdairi, M. Badr . Soft rough fuzzy sets based on covering. AIMS Mathematics, 2024, 9(5): 11180-11193. doi: 10.3934/math.2024548
[3]	Fahad Alsharari, Ahmed O. M. Abubaker, Islam M. Taha . On $r$ -fuzzy soft $\gamma$ -open sets and fuzzy soft $\gamma$ -continuous functions with some applications. AIMS Mathematics, 2025, 10(3): 5285-5306. doi: 10.3934/math.2025244
[4]	Abdelghani Taouti, Waheed Ahmad Khan . Fuzzy subnear-semirings and fuzzy soft subnear-semirings. AIMS Mathematics, 2021, 6(3): 2268-2286. doi: 10.3934/math.2021137
[5]	Arife Atay . Disjoint union of fuzzy soft topological spaces. AIMS Mathematics, 2023, 8(5): 10547-10557. doi: 10.3934/math.2023535
[6]	Samirah Alzahrani, A. A. Nasef, N. Youns, A. I. EL-Maghrabi, M. S. Badr . Soft topological approaches via soft γ-open sets. AIMS Mathematics, 2022, 7(7): 12144-12153. doi: 10.3934/math.2022675
[7]	Rui Gao, Jianrong Wu . Filter with its applications in fuzzy soft topological spaces. AIMS Mathematics, 2021, 6(3): 2359-2368. doi: 10.3934/math.2021143
[8]	Warud Nakkhasen, Teerapan Jodnok, Ronnason Chinram . Intra-regular semihypergroups characterized by Fermatean fuzzy bi-hyperideals. AIMS Mathematics, 2024, 9(12): 35800-35822. doi: 10.3934/math.20241698
[9]	Fenhong Li, Liang Kong, Chao Li . Non-global nonlinear mixed skew Jordan Lie triple derivations on prime $\ast$ -rings. AIMS Mathematics, 2025, 10(4): 7795-7812. doi: 10.3934/math.2025357
[10]	Umar Ishtiaq, Fahad Jahangeer, Doha A. Kattan, Manuel De la Sen . Generalized common best proximity point results in fuzzy multiplicative metric spaces. AIMS Mathematics, 2023, 8(11): 25454-25476. doi: 10.3934/math.20231299

Abstract

1. Introduction

In this paper, we consider the existence of solutions and a generalized Lyapunov-type inequality to the following boundary value problem for differential equation of variable order

$\begin{equation} \left\{ \begin{array}{ll} D_{0+}^{q(t)} x(t)+f(t, x) = 0, \ \ 0 \lt t \lt T, \\ x(0) = 0, \; \; x(T) = 0, \end{array}\right. \end{equation}$

(1.1)

where $0 < T < +\infty$ , $D_{0+}^{q(t)}$ denotes derivative of variable order(^[1,2,3,4]) defined by

$\begin{equation} D_{0+}^{q(t)}x(t) = \frac {d^2}{dt^2}\int_0^t\frac{(t-s)^{1-q(s)}}{\Gamma(2-q(s))}x(s)ds, \quad t \gt 0, \end{equation}$

(1.2)

and

$\begin{equation} I_{0+}^{2-q(t)}x(t) = \int_0^t\frac{(t-s)^{1-q(s)}}{\Gamma(2-q(s))}x(s)ds, \quad t \gt 0, \end{equation}$

(1.3)

denotes integral of variable order $2-q(t)$ , $1 < q(t)\leq 2$ , $0\leq t\leq T$ . $f:(0, T]\times\mathbb{R}\rightarrow \mathbb{R}$ is given continuous function satisfying some assumption conditions.

The operators of variable order, which fall into a more complex operator category, are the derivatives and integrals whose order is the function of certain variables. The variable order fractional derivative is an extension of constant order fractional derivative. In recent years, the operator and differential equations of variable order have been applied in engineering more and more frequently, for the examples and details, see ^{[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17]}.

The subject of fractional calculus has gained considerable popularity and importance due to its frequent appearance in different research areas and engineering, such as physics, chemistry, control of dynamical systems etc. Recently, many people paid attention to the existence and uniqueness of solutions to boundary value problems for fractional differential equations. Although the existing literature on solutions of boundary value problems of fractional order (constant order) is quite wide, few papers deal with the existence of solutions to boundary value problems of variable order. According to (1.2) and (1.3), it is obviously that when $q(t)$ is a constant function, i.e. $q(t)\equiv q$ ( $q$ is a finite positive constant), then $I_{0+}^{q(t)}, D_{0+}^{q(t)}$ are the usual Riemann-Liouville fractional integral and derivative ^[18].

The following properties of fractional calculus operators $D_{0+}^q$ , $I_{0+}^q$ play an important part in discussing the existence of solutions of fractional differential equations.

Proposition 1.1. ^[18] The equality $I_{0+}^\gamma I_{0+}^\delta f(t) = I_{0+}^{\gamma+\delta}f(t)$ , $\gamma > 0, \delta > 0$ holds for $f\in L(0, b), 0 < b < +\infty$ .

Proposition 1.2. ^[18] The equality $D_{0+}^\gamma I_{0+}^\gamma f(t) = f(t)$ , $\gamma > 0$ holds for $f\in L(0, b), 0 < b < +\infty$ .

Proposition 1.3. ^[18] Let $1 < \alpha\leq 2$ . Then the differential equation

$D_{0+}^\alpha f = 0$

has solutions

$f(t) = c_1t^{\alpha-1}+c_2t^{\alpha-2}, c_1, c_2\in R.$

Proposition 1.4. ^[18] Let $1 < \alpha\leq 2$ , $f(t)\in L(0, b)$ , $D_{0+}^\alpha f\in L(0, b)$ . Then the following equality holds

$I_{0+}^\alpha D_{0+}^\alpha f(t) = f(t)+c_1t^{\alpha-1}+c_2t^{\alpha-2}, c_1, c_2\in R.$

These properties play a very important role in considering the existence of the solutions of differential equations for the Riemnn-Liouville fractional derivative, for details, please refer to ^[18]. However, from ^[1,2,16], for general functions $h(t), g(t)$ , we notice that the semigroup property doesn't hold, i.e., $I_{a+}^{h(t)} I_{a+}^{g(t)}\neq I_{a+}^{h(t)+g(t)}$ . Thus, it brings us extreme difficulties, we can't get these properties like Propositions 1.1-1.4 for the variable order fractional operators (integral and derivative). Without these properties for variable order fractional derivative and integral, we can hardly consider the existence of solutions of differential equations for variable order derivative by means of nonlinear functional analysis (for instance, some fixed point theorems).

Let's take Proposition 1.1 for example. To begin with the simplest case,

Example 1.5. Let $p(t) = t$ , $q(t) = 1$ , $f(t) = 1, 0\leq t\leq 3$ . Now, we calculate $I_{0+}^{p(t)}I_{0+}^{q(t)}f(t)|_{t = 1}$ and $I_{0+}^{p(t)+q(t)}f(t)|_{t = 1}$ defined in $(1.3)$ .

$\begin{align*} I_{0+}^{p(t)}I_{0+}^{q(t)}f(t)|_{t = 1}& = \int_0^1\frac{(1-s)^{s-1}}{\Gamma(s)} \int_0^s\frac{(s-\tau)^{ 1-1}}{\Gamma(1)}d\tau ds = \int_0^1\frac{(1-s)^{s-1}s}{\Gamma(s)}ds\\ & \approx 0.472. \end{align*}$

and

$\begin{eqnarray*} I_{0+}^{p(t)+q(t)}f(t)|_{t = 1} = \int_0^1\frac{(1-s)^{s}} {\Gamma(s+1)}ds = \int_0^1\frac{(1-s)^{s}} {s\Gamma(s)}ds\approx 0.686. \end{eqnarray*}$

Therefore,

$I_{0+}^{p(t)}I_{0+}^{q(t)}f(t)|_{t = 1} \neq I_{0+}^{p(t)+q(t)}f(t)|_{t = 1}.$

As a result, the Propositions 1.2 and Propositions 1.4 do not hold for $D_{0^+}^{p(t)}$ and $I_{0^+}^{p(t)}$ , such as, for function $f\in L(0, T), 0 < p(t) < 1, 0\leq t\leq T$ , we get

$D_{0^+}^{p(t)}I_{0^+}^{p(t)}f(t) = D^1(I_{0^+}^{1-p(t)}I_{0^+}^{p(t)}f(t))\neq D^1I_{0^+}^{1-p(t)+p(t)}f(t) = f(t), \; t\in (0, T],$

since we know that $I_{0^+}^{1-p(t)}I_{0^+}^{p(t)}f(t)\neq I_{0^+}^{1-p(t)+p(t)}f(t)$ for general function $f$ .

Now, we can conclude that Propositions 1.1–1.4 do not hold for $D_{0+}^{q(t)}$ and $I_{0+}^{q(t)}$ .

So, one can not transform a differential equation of variable order into an equivalent interval equation without the Propositions 1.1–1.4. It is a difficulty for us in dealing with the boundary value problems of differential equations of variable order. Since the equations described by the variable order derivatives are highly complex, difficult to handle analytically, it is necessary and significant to investigate their solutions.

In ^[16], by means of Banach Contraction Principle, Zhang considered the uniqueness result of solutions to initial value problem of differential equation of variable order

$\begin{equation} \begin{cases} D_{0+}^{p(t)}x(t) = f(t, x), 0 \lt t \leq T, \\ x(0) = 0, \end{cases} \end{equation}$

(1.4)

where $0 < T < +\infty$ , $D_{0+}^{p(t)}$ denotes derivative of variable order $p(t)$ (^[1,2,3,4]) defined by

$\begin{equation} D_{0+}^{p(t)}x(t) = \frac {d}{dt}\int_0^t\frac{(t-s)^{-p(t)}} {\Gamma(1-p(t))}x(s)ds, t \gt 0. \end{equation}$

(1.5)

and $\frac 1{\Gamma(1-p(t))}\int_0^t(t-s)^{-p(t)}x(s)ds$ is integral of variable order $1-p(t)$ for function $x(t)$ . And $p:[0, T]\rightarrow (0, 1]$ is a piecewise constant function with partition $P = \{[0, T_1], (T_1, T_2], (T_2, T_3], \cdots$ , $(T_{N^{*}-1}, T]\}$ ( $N^{*}$ is a given natural number) of the finite interval $[0, T]$ , i.e.

$p(t) = \sum\limits_{k = 1}^{N^{*}}q_kI_k(t), t\in [0, T],$

where $0 < q_k\leq 1, k = 1, 2, \cdots, N^{*}$ are constants, and $I_k$ is the indicator of the interval $[T_{k-1}, T_k], k = 1, 2, \cdots, N^{*}$ (here $T_0 = 0, T_{N^{*}} = T$ ), that is $I_k = 1$ for $t\in [T_{k-1}, T_k]$ , $I_k = 0$ for elsewhere.

In ^[17], the authors studied the Cauchy problem for variable order differential equations with a piecewise constant order function^[19]. Inspired by these works, we will study the boundary value problem (1.1) for variable order differential equation with a piecewise constant order function $q(t)$ in this paper.

Lyapunov's inequality is an outstanding result in mathematics with many different applications, see ^{[20,21,22,23,24,25]} and references therein. The result, as proved by Lyapunov^[20] in 1907, asserts that if $h: [a, b]\rightarrow R$ is a continuous function, then a necessary condition for the boundary value problem

$\begin{equation} \begin{cases} y^{''}(t)+h(t)y(t) = 0, a \lt t \lt b, \\ y(a) = y(b) = 0, \end{cases} \end{equation}$

(1.6)

to have a nontrivial solution is given by

$\begin{equation} \int_a^b|h(s)|ds \gt \frac 4{b-a}, \end{equation}$

(1.7)

where $-\infty < a < b < +\infty$ .

Lyapunov's inequality has taken many forms, including versions in the context of fractional (noninteger order) calculus, where the second-order derivative in (1.6) is substituted by a fractional operator of order $\alpha$ ,

$\begin{equation} \begin{cases} D_{a+}^\alpha y(t)+h(t)y(t) = 0, a \lt t \lt b, \\ y(a) = y(b) = 0, \end{cases} \end{equation}$

(1.8)

where $D_{a+}^\alpha$ is the Riemann-Liouville derivative of order $\alpha \in(1, 2]$ and $h: [a, b]\rightarrow \mathbb{R}$ is a continuous function. If (1.8) has a nontrivial solution, then

$\begin{equation*} \int_a^b|h(s)|ds \gt \Gamma(\alpha)\left(\frac 4{b-a}\right)^{\alpha-1}. \end{equation*}$

A Lyapunov fractional inequality can also be obtained by considering the fractional derivative in in the sense of Caputo instead of Riemann-Liouville ^[22]. More recently, there are some results of Lyapunov type inequalities for fractional boundary value problems. see ^[23,24]. In ^[25], authors obtained a generalization of inequality to boundary value problem as following

$\begin{equation} \begin{cases} D_{a+}^\alpha y(t)+h(t)f(y) = 0, a \lt t \lt b, \\ y(a) = y(b) = 0, \end{cases} \end{equation}$

(1.9)

where $D_{a+}^\alpha$ is the Riemann-Liouville derivative, $1 < \alpha\leq 2$ , and $h: [a, b]\rightarrow \mathbb{R}$ is a Lebesgue integrable function. Under some assumptions on the nonlinear term $f$ , authors got a generalization of inequality to the boundary value problem (1.9).

$\begin{equation} \int_a^b|h(s)|ds \gt \frac{4^{\alpha-1\;\;}\Gamma(\alpha)\eta}{(b-a)^{\alpha-1\;\;}f(\eta)}\;\;, \end{equation}$

(1.10)

where $\eta$ is maximum value of nontrivial solution to the boundary value problem (1.9).

Motivated by ^{[21,22,23,24,25]} and the above results, we focus on a generalized Lyapunov-type inequality to the boundary value problem (1.1) under certain assumptions of nonlinear term.

The paper is organized as following. In Section 2, we provide some necessary definitions associated with the boundary value problem (1.1). In Section 3, we establish the existence of solutions for the boundary value problem (1.1) by using the Schauder fixed point theorem. In Section 4, we investigative the generalized Lyapunov-type inequalities to the boundary value problem (1.1). In section 5, we give some examples are presented to illustrate the main results.

2. Preliminaries

For the convenience of the reader, we present here some necessary definitions that will be used to prove our main results.

Definition 2.1. A generalized interval is a subset $I$ of $\mathbb{R}$ which is either an interval (i.e. a set of the form $[a, b], (a, b), [a, b)$ or $(a, b]$ ); a point $\{a\}$ ; or the empty set $\emptyset$ .

Definition 2.2. If $I$ is a generalized interval. A partition of $I$ is a finite set $P$ of generalized intervals contained in $I$ , such that every $x$ in $I$ lies in exactly one of the generalized intervals $J$ in $P$ .

Example 2.3. The set $P = \big\{\{1\}, (1, 6), [6, 7), \{7\}, (7, 8]\big\}$ of generalized intervals is a partition of ^[1,8].

Definition 2.4. Let $I$ be a generalized interval, let $f:I\rightarrow R$ be a function, and let $P$ a partition of $I$ . $f$ is said to be piecewise constant with respect to $P$ if for every $J\in P$ , $f$ is constant on $J$ .

Example 2.5. The function $f:[1,6]\rightarrow \mathbb{R}$ defined by

$f(x) = \begin{cases} 3, \ \ 1\leq x \lt 3, \\ 4, \ \ x = 3, \\ 5, \ \ 3 \lt x \lt 6, \\ 2, \ \ x = 6, \end{cases}$

is piecewise constant with respect to the partition $\big\{[1,3], \{3\}, (3, 6), \{6\}\big\}$ of $[1,6]$ .

The following example illustrates that the semigroup property of the variable order fractional integral doesn't holds for the piecewise constant functions $p(t)$ and $q(t)$ defined in the same partition of finite interval $[a, b]$ .

Example 2.6. Let $p(t) = \left\{ \begin{array}{ll} 3, \ \ 0\leq t\leq 1, \\ 2, \ \ 1 < t\leq 4, \end{array} \right.$ $q(t) = \left\{ \begin{array}{ll} 2, \ \ 0\leq t\leq 1, \\ 3, \ \ 1 < t\leq 4, \end{array} \right.$ and $f(t) = 1, 0\leq t\leq 4$ . We'll verify $I_{0+}^{p(t)}I_{0+}^{q(t)}f(t)|_{t = 2}\neq I_{0+}^{p(t)+q(t)}f(t)|_{t = 2}$ , here, the variable order fractional integral is defined in (1.3). For $1\leq t\leq 4$ , we have

$\begin{eqnarray*} &&I_{0+}^{p(t)}I_{0+}^{q(t)}f(t)\\ \\ = &&\int_0^1\frac{(t-s)^{p(s)-1}}{\Gamma(p(s))} \int_0^s\frac{(s-\tau)^{q(\tau)-1}}{\Gamma(q(\tau))}d\tau ds+\int_1^t\frac{(t-s)^{p(s)-1}}{\Gamma(p(s))} \int_0^s\frac{(s-\tau)^{q(\tau)-1}}{\Gamma(q(\tau))}d\tau ds\\ \\ = &&\int_0^1\frac{(t-s)^{2}}{\Gamma(3)} \int_0^s\frac{(s-\tau)}{\Gamma(2)}d\tau ds+\int_1^t\frac{(t-s)}{\Gamma(2)}[ \int_0^1\frac{(s-\tau)}{\Gamma(2)}d\tau+ \int_1^s\frac{(s-\tau)^{2}}{\Gamma(3)}d\tau] ds\\ \\ = &&\int_0^1\frac{(t-s)^2s^2}{2\Gamma(3)}ds+\int_1^t(t-s)[\frac{s^2}2-\frac{(s-1)^2}2+\frac{(s-1)^3}6] ds\\ \\ = &&\int_0^1\frac{(t-s)^2s^2}{2\Gamma(3)}ds+\frac 16\int_1^t(t-s)(s^3-3s^2+9s-4)ds, \end{eqnarray*}$

thus, we have

$\begin{eqnarray*} I_{0+}^{p(t)}I_{0+}^{q(t)}f(t)|_{t = 2} = &&\int_0^1\frac{(2-s)^2s^2}{2\Gamma(3)}ds+\frac 16\int_1^2(2-s)(s^3-3s^2+9s-4)ds\\ \\ = &&\frac2{15}+\frac{17}{60} = \frac 5{12} \end{eqnarray*}$

$\begin{eqnarray*} I_{0+}^{p(t)+q(t)}f(t)|_{t = 2} = && \int_0^2\frac{(2-s)^{p(s)+q(s)-1}}{\Gamma(p(s)+q(s))}ds\\ \\ = &&\int_0^1\frac{(2-s)^{3+2-1}}{\Gamma(3+2)}ds+\int_1^2\frac{(2-s)^{2+3-1}}{\Gamma(2+3)}ds\\ \\ = &&\frac{31}{120}+\frac 1{120} = \frac 4{15}. \end{eqnarray*}$

Therefore, we obtain

$I_{0+}^{p( t)}I_{0+}^{q(t)}f(t)|_{t = 2}\neq I_{0+}^{p(t)+q(t)}f(t)|_{t = 2},$

which implies that the semigroup property of the variable order fractional integral doesn't hold for the piecewise constant functions $p(t)$ and $q(t)$ defined in the same partition $[0, 1], (1, 4]$ of finite interval $[0, 4]$ .

3. Existence result of solutions

We need the following assumptions:

$(H_1)$ Let $n^{*}\in N$ be an integer, $P = \{[0, T_1], (T_1, T_2], (T_2, T_3], \cdots, (T_{n^{*}-1}, T]\}$ be a partition of the interval $[0, T]$ , and let $q(t):[0, T]\rightarrow (1, 2]$ be a piecewise constant function with respect to $P$ , i.e.,

$\begin{equation} q(t) = \sum\limits_{k = 1}^Nq_k{I}_k(t) = \begin{cases} q_1, \ \ 0\leq t\leq T_1, \\ q_2, \ \ T_1 \lt t\leq T_2, \\ \cdots, \ \ \cdots, \\ q_{n^{*}}, \ \ T_{n^{*}-1} \lt t\leq T_{n^{*}} = T, \end{cases} \end{equation}$

(3.1)

where $1 < q_k\leq 2\; (k = 1, 2, \cdots, n^{*}$ ) are constants, and ${I}_k$ is the indicator of the interval $[T_{k-1}, T_k]$ , $k = 1, 2, \cdots, n^{*}$ (here $T_0 = 0, T_{n^{*}} = T$ ), that is, ${I}_k(t) = 1$ for $t\in [T_{k-1}, T_k]$ and ${I}_k(t) = 0$ for elsewhere.

$(H_2)$ Let $t^rf:[0, T]\times \mathbb{R}\rightarrow \mathbb{R}$ be a continuous function ( $0\leq r < 1$ ), there exist constants $c_1 > 0, c_2 > 0$ , $0 < \gamma < 1$ such that

$t^r|f(t, x(t))|\leq c_1+c_2|x(t)|^\gamma, 0\leq t\leq T, x(t)\in \mathbb{R}.$

In order to obtain our main results, we firstly carry on essential analysis to the boundary value problem (1.1).

By (1.2), the equation of the boundary value problem (1.1) can be written as

$\begin{equation} \frac {d^2}{dt^2}\int_0^t\frac{(t-s)^{1-q(s)}} {\Gamma(2-q(s))}x(s)ds+f(t, x) = 0, \ \ 0 \lt t \lt T, \end{equation}$

(3.2)

According to $(H_1)$ , Eq (3.2) in the interval $(0, T_1]$ can be written as

$\begin{equation} D_{0+}^{q_1} x(t)+f(t, x) = 0, \ \ 0 \lt t\leq T_1. \end{equation}$

(3.3)

Equation (3.2) in the interval $(T_1, T_2]$ can be written by

$\begin{equation} \frac {d^2}{dt^2}\bigg(\int_0^{T_1}\frac{(t-s)^{1-q_1}} {\Gamma(2-q_1)}x(s)ds+\int_{T_1}^t\frac{(t-s)^{1-q_2}} {\Gamma(2-q_2)}x(s)ds\bigg)+f(t, x) = 0, \end{equation}$

(3.4)

and Eq (3.2) in the interval $(T_2, T_3]$ can be written by

$\begin{align} &\frac {d^2}{dt^2}\left(\int_0^{T_1}\frac{(t-s)^{1-q_1}} {\Gamma(2-q_1)}x(s)ds+\int_{T_1}^{T_2}\frac{(t-s)^{1-q_2}} {\Gamma(2-q_2)}x(s)ds +\int_{T_2}^t\frac{(t-s)^{1-q_3}} {\Gamma(2-q_3)}x(s)ds\right)\\ &+ f(t, x) = 0. \end{align}$

(3.5)

In the same way, Eq (3.2) in the interval $(T_{i-1}, T_i], \; i = 4, 5, \cdots, n^{*}-1$ can be written by

$\begin{equation} \frac {d^2}{dt^2}\left(\int_0^{T_1}\frac{(t-s)^{1-q_1}} {\Gamma(2-q_1)}x(s)ds+\cdots+\int_{T_{i-1}}^t\frac{(t-s)^{1-q_i}} {\Gamma(2-q_i)}x(s)ds\right)+f(t, x) = 0 . \end{equation}$

(3.6)

As for the last interval $(T_{n^{*}-1}, T)$ , similar to above argument, Eq (3.2) can be written by

$\begin{equation} \frac {d^2}{dt^2}\left(\int_0^{T_1}\frac{(t-s)^{1-q_1}} {\Gamma(2-q_1)}x(s)ds+\cdots+\int_{T_{n^{*}-1}}^t\frac{(t-s)^{1-q_{n^{*}}}} {\Gamma(2-q_{n^{*}})}x(s)ds\right)+f(t, x) = 0 . \end{equation}$

(3.7)

Remark 3.1. From the arguments above, we find that, according to condition $(H_1)$ , in the different interval, the equation of the boundary value problem (1.1) must be represented by different expression. For instance, in the interval $(0, T_1]$ , the equation of the boundary value problem (1.1) is represented by (3.3); in the interval $(T_1, T_2]$ , the equation of the boundary value problem (1.1) is represented by (3.4); in the interval $(T_2, T_3]$ , the equation of the boundary value problem (1.1) is represented by (3.5), etc. But, as far as we know, in the different intervals, the equation of integer order or constant fractional order problems may be represented by the same expression. Based these facts, different than integer order or constant fractional order problems, in order to consider the existence results of solution to the boundary value problem (1.1), we need consider the relevant problem defined in the different interval, respectively.

Now, based on arguments previous, we present definition of solution to the boundary value problem (1.1), which is fundamental in our work.

Definition 3.2. We say the boundary value problem (1.1) has a solution, if there exist functions $x_i(t), i = 1, 2, \cdots, n^{*}$ such that $x_1\in C[0, T_1]$ satisfying equation (3.3) and $x_1(0) = 0 = x_1(T_1)$ ; $x_2\in C[0, T_2]$ satisfying equation (3.4) and $x_2(0) = 0 = x_2(T_2)$ ; $x_3\in C[0, T_3]$ satisfying equation (3.5) and $x_3(0) = 0 = x_3(T_3)$ ; $x_i\in C[0, T_i]$ satisfying equation (3.6) and $x_i(0) = 0 = x_i(T_i)$ ( $i = 4, 5, \cdots, n^{*}-1$ ); $x_{n^{*}}\in C[0, T]$ satisfying equation (3.7) and $x_{n^{*}}(0) = x_{n^{*}}(T) = 0$ .

Theorem 3.3. Assume that conditions $(H_1)$ and $(H_2)$ hold, then the boundary value problem (1.1) has one solution.

Proof. According the above analysis, the equation of the boundary value problem (1.1) can be written as Eq (3.2). Equation (3.2) in the interval $(0, T_1]$ can be written as

$D_{0+}^{q_1} x(t)+f(t, x) = 0, \ \ 0 \lt t\leq T_1.$

Now, we consider the following two-point boundary value problem

$\begin{equation} \left\{ \begin{array}{ll} D_{0+}^{q_1} x(t)+f(t, x) = 0, \ \ 0 \lt t \lt T_1, \\ x(0) = 0, \; \; x(T_1) = 0. \end{array}\right. \end{equation}$

(3.8)

Let $x\in C[0, T_1]$ be solution of the boundary value problem (3.8). Now, applying the operator $I_{0+}^{q_1}$ to both sides of the above equation. By Propositions 1.4, we have

$x(t) = d_1t^{q_1-1}+d_2t^{q_1-2}-\frac 1{\Gamma(q_1)}\int_0^t(t-s)^{q_1-1}f(s, x(s))ds, \quad 0 \lt t\leq T_1.$

By $x(0) = 0$ and the assumption of function $f$ , we could get $d_2 = 0$ . Let $x(t)$ satisfying $x(T_1) = 0$ , thus we can get $d_1 = I_{0+}^{q_1}f(T_1, x)T_1^{1-q_1}$ . Then, we have

$\begin{eqnarray} x(t) = I_{0+}^{q_1}f(T_1, x)T_1^{1-q_1}t^{q_1-1}-I_{0+}^{q_1}f(t, x), 0\leq t\leq T_1 \end{eqnarray}$

(3.9)

Conversely, let $x\in C[0, T_1]$ be solution of integral Eq (3.9), then, by the continuity of function $t^rf$ and Proposition 1.2, we can easily get that $x$ is the solution of boundary value problem (3.8).

Define operator $T:C[0, T_1]\rightarrow C[0, T_1]$ by

$\begin{equation*} Tx(t) = I_{0+}^{q_1}f(T_1, x)T_1^{1-q_1} t^{q_1-1}-I_{0+}^{q_1}f\big(t, x(t)), \quad 0\leq t\leq T_1. \end{equation*}$

It follows from the properties of fractional integrals and assumptions on function $f$ that the operator $T:C[0, T_1]\rightarrow C[0, T_1]$ defined above is well defined. By the standard arguments, we could verify that $T:C[0, T_1]\rightarrow C[0, T_1]$ is a completely continuous operator.

In the next analysis, we take

$M(r, q) = \max\left\{\frac 2{(1-r)\Gamma(q_1)}, \frac 2{(1-r)\Gamma(q_2)}, \cdots, \frac 2{(1-r)\Gamma(q_{n^{*}})}\right\}.$

Let $\Omega = \{x\in C[0, T_1]: \|x\|\leq R\}$ be a bounded closed convex subset of $C[0, T_1]$ , where

$R = \max\left\{ 2c_1M(r, q)(T+1)^2, \left(2c_2M(r, q)(T+1)^2\right)^{\frac 1{1-\gamma}}\right \}.$

For $x\in \Omega$ and by $(H_2)$ , we have

$\begin{eqnarray*} |Tx(t)|\leq&&\frac {T_1^{1-q_1}t^{q_1-1}}{\Gamma(q_1)}\int_0^{T_1}(T_1-s)^{q_1-1}|f(s, x(s))|ds+\frac 1{\Gamma(q_1)}\int_0^t(t-s)^{q_1-1}|f(s, x(s))|ds\\ \\ \leq&&\frac 2{\Gamma(q_1)}\int_0^{T_1}(T_1-s)^{q_1-1}|f(s, x(s))|ds\\ \\ \leq&&\frac 2{\Gamma(q_1)}\int_0^{T_1}(T_1-s)^{q_1-1}s^{-r}(c_1+c_2|x(s)|^{\gamma}) ds\\ \\ \leq&&\frac {2T_1^{q_1-1}}{\Gamma(q_1)}\int_0^{T_1}s^{-r}(c_1+c_2R^{\gamma})ds\\ \\ \leq&&\frac{ 2T_1^{q_1-1}T_1^{1-r}}{(1-r)\Gamma(q_1)}(c_1+c_2R^{\gamma})\\ \\ \leq&&M(r, q)T_1^{q_1-r}(c_1+c_2R^{\gamma})\\ \\ \leq&&M(r, q)(T+1)^2(c_1+c_2RR^{\gamma-1})\\ \\ \leq&&\frac R2+\frac R2 = R, \end{eqnarray*}$

which means that $T(\Omega)\subseteq \Omega$ . Then the Schauder fixed point theorem assures that the operator $T$ has one fixed point $x_1\in \Omega$ , which is a solution of the boundary value problem (3.8).

Also, we have obtained that Eq (3.2) in the interval $(T_1, T_2]$ can be written by (3.4). In order to consider the existence result of solution to (3.4), we rewrite (3.4) as following

$\begin{equation*} \frac {d^2}{dt^2}\int_0^{T_1}\frac{(t-s)^{1-q_2}} {\Gamma(2-q_2)}x(s)ds+\frac {d^2}{dt^2}\int_{T_1}^t\frac{(t-s)^{1-q_2}} {\Gamma(2-q_2)}x(s)ds = f(t, x).\ \ T_1 \lt t\leq T_2, \end{equation*}$

For $0\leq s\leq T_1$ , we take $x(s)\equiv 0$ , then, by the above equation, we get

$D_{T_1}^{q_2}x(t)+f(t, x) = 0, T_1 \lt t \lt T_2.$

Now, we consider the following boundary value problem

$\begin{equation} \left\{ \begin{array}{ll} D_{T_1+}^{q_2} x(t)+f(t, x) = 0, \ \ T_1 \lt t \lt T_2, \\ x(T_1) = 0, \; \; x(T_2) = 0, \end{array}\right. \end{equation}$

(3.10)

Let $x\in C[T_1, T_2]$ be solution of the boundary value problem (3.10). Now, applying operator $I_{T_1+}^{q_2}$ on both sides of equation to boundary value problem (3.10) and by Propositions 1.4, we have

$x(t) = d_1(t-T_1)^{q_2-1}+d_2(t-T_1)^{q_2-2}-\frac 1{\Gamma(q_2)} \int_{T_1}^t(t-s)^{q_2-1}f(s, x(s))ds, \; \; T_1 \lt t\leq T_2.$

By $x(T_1) = 0, x(T_2) = 0$ , we have $d_2 = 0$ and $d_1 = I_{T_1+}^{q_2}f(T_2, x)(T_2-T_1)^{1-q_2}$ . Then, we have

$\begin{align*} x(t) = I_{0+}^{q_2}f(T_2, x)(T_2&-T_1)^{1-q_2}(t-T_1)^{q_2-1}- \frac 1{\Gamma(q_2)}\int_{T_1}^t(t-s)^{q_2-1}f(s, x(s))ds, \; \; T_1\leq t\leq T_2. \end{align*}$

Conversely, let $x\in C[T_1, T_2]$ be solution of integral equation above, then, by the continuity assumption of function $t^rf$ and Proposition (1.2), we can get that $x$ is solution solution of the boundary value problem (3.10).

Define operator $T:C[T_1, T_2]\rightarrow C[T_1, T_2]$ by

$Tx(t) = I_{0+}^{q_2}f(T_2, x)(T_2-T_1)^{1-q_2}(t-T_1)^{q_2-1}- \frac 1{\Gamma(q_2)}\int_{T_1}^t(t-s)^{q_2-1}f(s, x(s))ds.$

It follows from the continuity of function $t^rf$ that operator $T:C[T_1, T_2]\rightarrow C[T_1, T_2]$ is well defined. By the standard arguments, we know that $T:C[T_1, T_2]\rightarrow C[T_1, T_2]$ is a completely continuous operator.

For $x\in \Omega$ and by $(H_2)$ , we get

$\begin{eqnarray*} |Tx(t)|\leq&&\frac {(T_2-T_1)^{1-q_2}(t-T_1)^{q_2-1}}{\Gamma(q_2)}\int_{T_1}^{T_2}(T_2-s)^{q_2-1}|f(s, x(s))|ds+\frac 1{\Gamma(q_2)}\int_{T_1}^t(t-s)^{q_2-1}|f(s, x(s))|ds\\ \\ \leq&&\frac 2{\Gamma(q_2)}\int_{T_1}^{T_2}(T_2-s)^{q_2-1}|f(s, x(s))|ds\\ \\ \leq&&\frac 2{\Gamma(q_2)}\int_{T_1}^{T_2}(T_2-s)^{q_2-1}s^{-r}(c_1+c_2|x(s)|^{\gamma}) ds\\ \\ \leq&&\frac {2T_2^{q_2-1}}{\Gamma(q_2)}\int_{T_1}^{T_2}s^{-r}(c_1+c_2R^{\gamma}) ds\\ \\ = &&\frac{2T_2^{q_2-1} (T_2^{1-r}-T_1^{1-r})} {(1-r)\Gamma(q_2)}(c_1+c_2R^{\gamma})\\ \\ \leq&&\frac{2T_2^{q_2-r}} {(1-r)\Gamma(q_2)}(c_1+c_2R^{\gamma})\\ \\ \leq&&M(r, q)(T+1)^2(c_1+c_2RR^{\gamma-1})\\ \\ \leq&&\frac R2+\frac R2 = R, \end{eqnarray*}$

which means that $T(\Omega)\subseteq \Omega$ . Then the Schauder fixed point theorem assures that operator $T$ has one fixed point $\widetilde{x}_2\in \Omega$ , which is one solution of the following integral equation, that is,

$\begin{align} \widetilde{x}_2(t)& = I_{0+}^{q_2}f(T_2, \widetilde{x}_2)(T_2-T_1)^{1-q_2} (t-T_1)^{q_2-1}\\ &-\frac 1{\Gamma(q_2)}\int_{T_1}^t(t-s)^{q_2-1} f(s, \widetilde{x}_2(s))ds, \quad T_1\leq t\leq T_2. \end{align}$

(3.11)

Applying operator $D_{T_1+}^{q_2}$ on both sides of (3.11), by Proposition $1.2$ , we can obtain that

$D_{T_1+}^{q_2}\widetilde{x}_2(t)+f(t, \widetilde{x}_2) = 0, \quad T_1 \lt t\leq T_2,$

that is, $\widetilde{x}_2(t)$ satisfies the following equation

$\begin{equation} \frac {d^2}{dt^2}\frac 1{\Gamma(2-q_2)} \int_{T_1}^t(t-s)^{1-q_2}\widetilde{x}_2(s)ds+f(t, \widetilde{x}_2) = 0, \quad T_1 \lt t\leq T_2. \end{equation}$

(3.12)

We let

$\begin{equation} x_2(t) = \left\{ \begin{array}{ll} 0, \quad\quad \ \ 0\leq t\leq T_1, \\ \widetilde{x}_2(t), \quad T_1 \lt t\leq T_2 \end{array}\right. \end{equation}$

(3.13)

hence, from (3.12), we know that $x_2\in C[0, T_2]$ defined by (3.13) satisfies equation

$\begin{eqnarray*} \frac {d^2}{dt^2}\left (\int_0^{T_1}\frac{(t-s)^{1-q_1}} {\Gamma(2-q_1)}x_2(s)ds+\int_{T_1}^t\frac{(t-s)^{1-q_2}} {\Gamma(2-q_2)}x_2(s)ds\right)+f(t, x_2) = 0, \end{eqnarray*}$

which means that $x_2\in C[0, T_2]$ is one solution of (3.4) with $x_2(0) = 0, x_2(T_2) = \widetilde{x}_2(T_2) = 0$ .

Again, we have known that Eq (3.2) in the interval $(T_2, T_3]$ can be written by (3.5). In order to consider the existence result of solution to Eq (3.5), for $0\leq s\leq T_2$ , we take $x(s)\equiv 0$ , then, by (3.5), we get

$D_{T_2}^{q_3}x(t)+f(t, x) = 0, \; \; T_2 \lt t \lt T_3.$

Now, we consider the following boundary value problem

$\begin{equation} \left\{ \begin{array}{ll} D_{T_2+}^{q_3} x(t)+f(t, x) = 0, \ \ T_2 \lt t \lt T_3, \\ x(T_2) = 0, \; \; x(T_3) = 0. \end{array}\right. \end{equation}$

(3.14)

By the standard way, we know that the boundary value problem (3.14) exists one solution $\widetilde{x}_3\in \Omega$ . Since $\widetilde{x}_3$ satisfies equation

$D_{T_2+}^{q_3}\widetilde{x}_3(t)+f(t, \widetilde{x}_3) = 0, \quad T_2 \lt t\leq T_3,$

that is, $\widetilde{x}_3(t)$ satisfies the following equation

$\begin{equation} \frac {d^2}{dt^2}\frac 1{\Gamma(2-q_3)} \int_{T_2}^t(t-s)^{1-q_3}\widetilde{x}_3(s)ds+f(t, \widetilde{x}_3) = 0, \quad T_2 \lt t\leq T_3. \end{equation}$

(3.15)

We let

$\begin{equation} x_3(t) = \left\{ \begin{array}{ll} 0, \quad\quad \ \ 0\leq t\leq T_2, \\ \widetilde{x}_3(t), \quad \quad T_2 \lt t\leq T_3, \end{array}\right. \end{equation}$

(3.16)

hence, from (3.15), we know that $x_3\in C[0, T_3]$ defined by (3.16) satisfies equation

$\begin{align*} \frac {d^2}{dt^2}\bigg(\int_0^{T_1}\frac{(t-s)^{1-q_1}}{\Gamma(2-q_1)}x_3(s)ds &+ \int_{T_1}^{T_2}\frac{(t-s)^{1-q_2}} {\Gamma(2-q_2)}x_3(s)ds\\ &+ \int_{T_2}^t\frac{(t-s)^{1-q_3}}{\Gamma(2-q_3)}x_3(s)ds\bigg)+f(t, x_3) = 0, \end{align*}$

which means that $x_3\in C[0, T_3]$ is one solution of (3.5) with $x_3(0) = 0, x(T_3) = \widetilde{x}_3(T_3) = 0$ .

By the similar way, in order to consider the existence of solution to Eq (3.6) defined on $[T_{i-1}, T_i]$ of (3.2), we can investigate the following two-point boundary value problem

$\begin{equation} \left\{ \begin{array}{ll} D_{T_{i-1}+}^{q_i} x(t)+f(t, x) = 0, \ \ T_{i-1} \lt t \lt T_i, \\ x(T_{i-1}) = 0, \; \; x(T_i) = 0. \end{array}\right. \end{equation}$

(3.17)

By the same arguments previous, we obtain that the Eq (3.6) defined on $[T_{i-1}, T_i]$ of (3.2) has solution

$\begin{equation} x_i(t) = \left\{ \begin{array}{ll} 0, \quad\quad \ \ 0\leq t\leq T_{i-1}, \\ \widetilde{x}_i(t), \quad T_{i-1} \lt t\leq T_i, \end{array}\right. \end{equation}$

(3.18)

where $\widetilde{x}_i\in \Omega$ with $\widetilde{x}_i(T_{i-1}) = 0 = \widetilde{x}_i(T_i)$ , $i = 4, 5, \cdots, n^{*}-1$ .

Similar to the above argument, in order to consider the existence result of solution to Eq (3.7), we may consider the following boundary value problem

$\begin{equation} \left\{ \begin{array}{ll} D_{T_{n^{*}-1}+}^{q_{n^{*}}} x(t)+f(t, x) = 0, \ \ T_{n^{*}-1} \lt t \lt T_{n^{*}} = T, \\ x(T_{n^{*}-1}) = 0, \; \; x(T) = 0. \end{array}\right. \end{equation}$

(3.19)

So by the same considering, for $T_{n^{*}-1}\leq t\leq T$ we get

$\begin{eqnarray*} x(t) = (T-T_{n^{*}-1})^{1-q_{n^{*}}}(t-T_{n^{*}-1})^ {q_{n^{*}}-1}I_{T_{n^{*}-1}+}^{q_{n^{*}}} f(T, x)-I_{T_{n^{*}-1}+}^{q_{n^{*}}}f(t, x). \end{eqnarray*}$

Define operator $T:C[T_{n^{*}-1}, T]\rightarrow C[T_{n^{*}-1}, T]$ by

$\begin{align*} Tx(t)& = (T-T_{n^{*}-1})^{1-q_{n^{*}}}(t-T_{n^{*}-1})^{q_{n^{*}}-1} I_{T_{n^{*}-1}+}^{q_{n^{*}}}f(T, x)-\frac 1{\Gamma(q_{n^{*}})} \int_{T_{n^{*}-1}}^t(t-s)^{q_{n^{*}}-1}f(s, x(s))ds, \end{align*}$

$T_{n^{*}-1}\leq t\leq T.$ It follows from the continuity assumption of function $t^rf$ that operator $T:C[T_{n^{*}-1}, T]\rightarrow C[T_{n^{*}-1}, T]$ is well defined. By the standard arguments, we note that $T:C[T_{n^{*}-1}, T]\rightarrow C[T_{n^{*}-1}, T]$ is a completely continuous operator.

For $x\in \Omega$ and by $(H_2)$ , we get

$\begin{eqnarray*} |Tx(t)|\leq&&\frac {(T-T_{n^{*}-1})^{1-q_{n^{*}}}(t-T_{n^{*}-1})^{q_{n^{*}}-1}}{\Gamma(q_{n^{*}})}\int_{T_{n^{*}-1}}^{T}(T-s)^{q_{n^{*}}-1}|f(s, x(s))|ds\\ \\ &&+\frac 1{\Gamma(q_{n^{*}})} \int_{T_{n^{*}-1}}^t(t-s)^{q_{n^{*}}-1}|f(s, x(s))|ds\\ \\ \leq&&\frac 2{\Gamma(q_{n^{*}})} \int_{T_{n^{*}-1}}^T(T-s)^{q_{n^{*}}-1}|f(s, x(s))|ds\\ \\ \leq&&\frac 2{\Gamma(q_{n^{*}})} \int_{T_{n^{*}-1}}^T(T-s)^{q_{n^{*}}-1}s^{-r}(c_1+c_2|x(s)|^{\gamma}) ds\\ \\ \leq&&\frac{ 2T^{q_{n^{*}}-1}}{\Gamma(q_{n^{*}})} \int_{T_{n^{*}-1}}^Ts^{-r}(c_1+c_2R^{\gamma}) ds\\ \\ \leq&&\frac{ 2T^{q_{n^{*}}-1}(T^{1-r}-T_{n^{*}-1}^{1-r})}{(1-r)\Gamma(q_{n^{*}})}(c_1+c_2R^{\gamma})\\ \\ \leq&&\frac{ 2(T+1)^2}{(1-r)\Gamma(q_{n^{*}})}(c_1+c_2R^{\gamma})\\ \\ \leq&&M(r, q)(T+1)^2(c_1+c_2RR^{\gamma-1})\\ \\ \leq&&\frac R2+\frac R2 = R, \end{eqnarray*}$

which means that $T(\Omega)\subseteq \Omega$ . Then the Schauder fixed point theorem assures that operator $T$ has one fixed point $\widetilde{x}_{n^{*}}\in \Omega$ , which is one solution of the following integral equation, that is,

$\begin{align} \widetilde{x}_{n^{*}}(t)& = (T-T_{n^{*}-1})^{1-q_{n^{*}}} (t-T_{n^{*}-1})^{q_{n^{*}}-1} I_{T_{n^{*}-1}+}^{q_{n^{*}}}f(T, \widetilde{x}_{n^{*}}) \\ &-\frac 1{\Gamma(q_{n^{*}})}\int_{T_{n^{*}-1}}^t(t-s)^{q_{n^{*}}-1} f(s, \widetilde{x}_{n^{*}}(s))ds, T_{n^{*}-1}\leq t\leq T. \end{align}$

(3.20)

Applying operator $D_{T_{n^{*}-1}+}^{q_{n^{*}}}$ on both sides of (3.20), by Proposition 1.2, we can obtain that

$D_{T_{n^{*}-1}+}^{q_{n^{*}}}\widetilde{x}_{n^{*}}(t)+f(t, \widetilde{x}_{n^{*}}) = 0, \quad T_{n^{*}-1} \lt t\leq T,$

that is, $\widetilde{x}_T(t)$ satisfies the following equation

$\begin{equation} \frac {d^2}{dt^2}\frac 1{\Gamma(2-q_{n^{*}})} \int_{T_{n^{*}-1}}^t(t-s)^{1-q_{n^{*}}}\widetilde{x}_{n^{*}}(s)ds+f(t, \widetilde{x}_{n^{*}}) = 0, \quad T_{n^{*}-1} \lt t\leq T. \end{equation}$

(3.21)

We let

$\begin{equation} x_{n^{*}}(t) = \left\{ \begin{array}{ll} 0, \quad\quad \ \ 0\leq t\leq T_{n^{*}-1}, \\ \widetilde{x}_{n^{*}}(t), \quad T_{n^{*}-1} \lt t\leq T, \end{array}\right. \end{equation}$

(3.22)

hence, from (3.21), we know that $x_{n^{*}}\in C[0, T]$ defined by (3.22) satisfies equation

$\begin{align*} \frac {d^2}{dt^2}\bigg(\int_0^{T_1}\frac{(t-s)^{1-q_1}} {\Gamma(2-q_1)}&x_{n^{*}}(s)ds+\cdots\\&+\int_{T_{n^{*}-1}}^t\frac{(t-s)^{1-q_{n^{*}}}} {\Gamma(2-q_{n^{*}})}x_{n^{*}}(s)ds\bigg)+f(t, x_{n^{*}}) = 0. \end{align*}$

for $T_{n^{*}-1} < t < T$ , which means that $x_{n^{*}}\in C[0, T]$ is one solution of (3.7) with $x_{n^{*}}(0) = 0, x_{n^{*}}(T) = \widetilde{x}_{n^{*}}(T) = 0$ .

As a result, we know that the boundary value problem (1.1) has a solution. Thus we complete the proof.

Remark 3.4. For condition $(H_2)$ , if $\gamma\geq 1$ , then using similar way, we can obtain the existence result of solution to the boundary value problem (1.1) provided that we impose some additional conditions on $c_1, c_2$ .

4. The generalized Lyapunov-type inequalities

In this section, we investigate the generalized Lyapunov-type inequalities for the boundary value problem (1.1).

Now, we explore characters of Green functions to the boundary value problems (3.8), (3.10), (3.14), $\cdots$ , (3.17) and (3.19).

Proposition 4.1. Assume that $t^rf:[0, T]\times \mathbb{R}\rightarrow \mathbb{R}$ , $(0\leq r < 1)$ is continuous function, $q(t):[0, T]\rightarrow (1, 2]$ satisfies $(H_1)$ , then the Green functions

$\begin{equation} G_i(t, s) = \begin{cases} \frac 1{\Gamma(q_i)}[(T_i-T_{i-1})^{1-q_i}(t-T_{i-1})^{q_i-1} (T_i-s)^{q_i-1}-(t-s)^{q_i-1}], \\ \qquad\qquad T_{i-1}\leq s\leq t\leq T_i, \\ \frac 1{\Gamma(q_i)}(T_i-T_{i-1})^{1-q_i}(t-T_{i-1})^{q_i-1}(T_i-s)^{q_i-1}, \\ \qquad\qquad T_{i-1}\leq t\leq s\leq T_i, \end{cases} \end{equation}$

(4.1)

of the boundary value problems (3.8), (3.10), (3.14), $\cdots$ , (3.17) and (3.19) satisfy the following properties:

$(1)$ $G_i(t, s)\geq 0$ for all $T_{i-1}\leq t, s\leq T_i$ ;

$(2)$ $\max_{t\in [T_{i-1}, T_i]}G_i(t, s) = G_i(s, s)$ , $s\in [T_{i-1}, T_i]$ ;

$(3)$ $G_i(s, s)$ has one unique maximum given by

$\max\limits_{s\in [T_{i-1}, T_i]}G_i(s, s) = \frac 1{\Gamma(q_i)}(\frac{T_i-T_{i-1}}4)^{q_i-1},$

where $i = 1, 2, \cdots, n^{*}$ , $T_0 = 0, T_{n^{*}} = T$ .

Proof. From the proof of Theorem 3.1, we know that Green functions of the boundary value problems (3.8), (3.10), (3.14), $\cdots$ , (3.17) and (3.19) are given by (4.1).

Using a similar way, we can verify these three results. In fact, let

$g(t, s) = (T_i-T_{i-1})^{1-q_i}(t-T_{i-1})^{q_i-1}(T_i-s)^{q_i-1}-(t-s)^{q_i-1}, T_{i-1}\leq s\leq t\leq T_i.$

We see that

$\begin{eqnarray*} g_t(t, s)&& = (q_i-1)[(T_i-T_{i-1})^{1-q_i}(t-T_{i-1})^{q_i-2}(T_i-s)^{q_i-1}-(t-s)^{q_i-2}]\\ \\ &&\leq(q_i-1)[(T_i-T_{i-1})^{1-q_i}(t-s)^{q_i-2}(T_i-T_{i-1})^{q_i-1}-(t-s)^{q_i-2}]\\ && = 0, \end{eqnarray*}$

which means that $g(t, s)$ is nonincreasing with respect to $t$ , so $g(t, s)\geq g(T_i, s) = 0$ for $T_{i-1}\leq s\leq t\leq T_i$ . Thus, together this with the expression of $G_i(t, s)$ , we get that $G_i(t, s)\geq 0$ for all $T_{i-1}\leq t, s\leq T_i$ , $i = 1, 2, \cdots, n^{*}$ , $T_0 = 0, T_{n^{*}} = T$ .

Since $g(t, s)$ is nonincreasing with respect to $t$ , it holds that $g(t, s)\leq g(s, s)$ for $T_{i-1}\leq s\leq t\leq T_i$ . On the other hand, for $T_{i-1}\leq t\leq s\leq T_i$ , we have

$(T_i-T_{i-1})^{1-q_i}(t-T_{i-1})^{q_i-1}(T_i-s)^{q_i-1}\leq (T_i-T_{i-1})^{1-q_i}(s-T_{i-1})^{q_i-1}(T_i-s)^{q_i-1}.$

These assure that $\max_{t\in [T_{i-1}, T_i]}G_i(t, s) = G_i(s, s)$ , $s\in [T_{i-1}, T_i]$ , $i = 1, 2, \cdots, n^{*}$ , $T_0 = 0, T_{n^{*}} = T$ .

Next, we verify $(3)$ of Proposition 4.1. Obviously, the maximum points of $G_i(s, s)$ are not $T_{i-1}$ and $T_i$ , $i = 1, 2, \cdots, n^{*}$ , $T_0 = 0, T_{n^{*}} = T$ . For $s\in (T_{i-1}, T_i)$ , $i = 1, 2, \cdots, n^{*}$ , $T_0 = 0, T_{n^{*}} = T$ , we have that

$\begin{align*} \frac{d G_i(s, s)}{ds}& = \frac 1{\Gamma(q_i)}(T_i-T_{i-1})^{1-q_i}(q_i-1)[(s-T_{i-1})^{q_i-2}(T_i-s)^{q_i-1}\\ &\qquad- (s-T_{i-1})^{q_i-1}(T_i-s)^{q_i-2}]\\ = &\frac 1{\Gamma(q_i)}(T_i-T_{i-1})^{1-q_i}(q_i-1)(s-T_{i-1})^{q_i-2}(T_i-s)^{q_i-2}[T_i+T_{i-1}-2s], \end{align*}$

which implies that the maximum points of $G_i(s, s)$ is $s = \frac{T_{i-1}+T_i}2$ , $i = 1, 2, \cdots, n^{*}$ , $T_0 = 0, T_{n^{*}} = T$ . Hence, for $i = 1, 2, \cdots, n^{*}, T_0 = 0, T_{n^{*}} = T$ ,

$\max\limits_{s\in [T_{i-1}, T_i]}G_i(s, s) = G\bigg(\frac{T_{i-1}+T_i}2, \frac{T_{i-1}+T_i}2\bigg) = \frac 1{\Gamma(q_i)}(\frac{T_i-T_{i-1}}4)^{q_i-1} .$

Thus, we complete this proof.

Theorem 4.2. Let $(H_1)$ holds and $t^rf:[0, T]\times \mathbb{R}\rightarrow \mathbb{R}$ , ( $0\leq r < 1$ ) be a continuous function. Assume that there exists nonnegative continuous function $h(t)$ defined on $[0, T]$ such that

$t^r|f(t, x)|\leq h(t)|x(t)|, 0\leq t\leq T, x(t)\in R$

If the boundary value problem (1.1) has a nontrivial solution $x$ , then

$\begin{equation} \int_{0}^{T}s^{-r}h(s)ds \gt \sum\limits_{i = 1}^{n^{*}}\Gamma(q_i)\bigg(\frac{4}{T_i-T_{i-1}}\bigg)^{q_i-1}. \end{equation}$

(4.2)

Proof. Let $x$ be a nontrivial solution of the boundary value problem (1.1). Using Definition 3.2 and the proof of Theorem 3.3, we know that

$\begin{equation} x(t) = \begin{cases} x_1(t), \quad 0\leq t\leq T_1\\ x_2(t) = \begin{cases} 0, \quad 0\leq t\leq T_1, \\ \widetilde{x}_1(t), \quad T_1 \lt t\leq T_2, \\ \end{cases} \\ \\ x_3(t) = \begin{cases} 0, \quad 0\leq t\leq T_2, \\ \widetilde{x}_2(t), \quad T_2 \lt t\leq T_3, \\ \end{cases} \\ \vdots\\ x_i(t) = \begin{cases} 0, \quad 0\leq t\leq T_{i-1}, \\ \widetilde{x}_i(t), \quad T_{i-1} \lt t\leq T_i, \\ \end{cases} \\ \vdots\\ x_{n^{*}}(t) = \begin{cases} 0, \quad 0\leq t\leq T_{n^{*}-1}, \\ \widetilde{x}_{n^{*}}(t), \quad T_{n^{*}-1} \lt t\leq T, \end{cases} \end{cases} \end{equation}$

(4.3)

where $x_1\in C[0, T_1]$ is nontrivial solution of the boundary value problem (3.8) with $a_1 = 0$ , $\widetilde{x}_2\in C[T_1, T_2]$ is nontrivial solution of the boundary value problem (3.10) with $a_2 = 0$ , $\widetilde{x}_3\in C[T_2, T_3]$ is nontrivial solution of the boundary value problem (3.14) with $a_3 = 0$ , $\widetilde{x}_i\in C[T_{i-1}, T_i]$ is nontrivial solution of the boundary value problem (3.17) with $a_i = 0$ , $\widetilde{x}_{n^{*}}\in C[T_{n^{*}-1}, T]$ is nontrivial solution of the boundary value problem (3.19). From (4.3) and Proposition 4.1, we have

$\begin{eqnarray*} \|x_1\|_{C[0, T_1]} = \max\limits_{0\leq t\leq T_1}|x_1(t)|\leq &&\max\limits_{0\leq t\leq T_1}\int_0^{T_1}G_1(t, s)|f(s, x_1(s))|ds\\ \\ \leq&& \int_0^{T_1}G_1(s, s)s^{-r}h(s)|x_1(s)|ds\\\\ \\ \lt &&\frac{\|x_1\|_{C[0, T_1]}}{\Gamma(q_1)}(\frac{T_1}4)^{q_1-1}\int_0^{T_1}s^{-r}h(s)ds, \ \end{eqnarray*}$

which implies that

$\begin{equation} \int_0^{T_1}s^{-r}h(s)ds \gt \Gamma(q_1)(\frac 4{T_1})^{q_1-1}. \end{equation}$

(4.4)

$\begin{eqnarray*} \|x_2\|_{C[0, T_2]} = \max\limits_{T_1\leq t\leq T_2}|\widetilde{x}_2(t)| = &&\max\limits_{T_1\leq t\leq T_2}\bigg|\int_{T_1}^{T_2}G_2(t, s)f(s, \widetilde{x}_2(s))ds\bigg|\\ \\ \leq&& \int_{T_1}^{T_2}G_2(s, s)s^{-r}h(s)|\widetilde{x}_2(s)|ds\\ \\ \lt &&\frac{\|\widetilde{x}_2\|_{C[T_1, T_2]}}{\Gamma(q_2)}(\frac{T_2-T_1}4)^{q_2-1}\int_{T_1}^{T_2}s^{-r}h(s)ds, \\ \\ = &&\frac{\|x_2\|_{C[0, T_2]}}{\Gamma(q_2)}(\frac{T_2-T_1}4)^{q_2-1}\int_{T_1}^{T_2}s^{-r}h(s)ds, \end{eqnarray*}$

which implies that

$\begin{equation} \int_{T_1}^{T_2}s^{-r}h(s)ds \gt \Gamma(q_2)(\frac 4{T_2-T_1})^{q_2-1}. \end{equation}$

(4.5)

Similar, for $i = 3, 4, \cdots, n^{*}$ ( $T_{n^{*}} = T)$ , we have

$\begin{eqnarray*} \|x_i\|_{C[0, T_i]}&& = \max\limits_{T_{i-1}\leq t\leq T_i}|\widetilde{x}_i(t)| = \max\limits_{T_{i-1}\leq t\leq T_i}\bigg|\int_{T_{i-1}}^{T_i}G_i(t, s)f(s, \widetilde{x}_i(s))ds\bigg|\\ \\ \leq&& \int_{T_{i-1}}^{T_i}G_i(s, s)s^{-r}h(s)|\widetilde{x}_i(s)|ds\\ \\ \lt &&\frac{\|\widetilde{x}_i\|_{C[T_{i-1}, T_i]}}{\Gamma(q_i)}(\frac{T_i-T_{i-1}}4)^{q_i-1}\int_{T_{i-1}}^{T_i}s^{-r}h(s)ds, \\ \\ = &&\frac{\|x_i\|_{C[0, T_i]}}{\Gamma(q_i)}(\frac{T_i-T_{i-1}}4)^{q_i-1}\int_{T_{i-1}}^{T_i}s^{-r}h(s)ds, \end{eqnarray*}$

which implies that

$\begin{equation*} \int_{T_{i-1}}^{T_i}s^{-r}h(s)ds \gt \Gamma(q_i)\bigg(\frac 4{T_i-T_{i-1}}\bigg)^{q_i-1}. \end{equation*}$

So, we get

$\begin{equation*} \int_{0}^{T}s^{-r}h(s)ds \gt \sum\limits_{i = 1}^{n^{*}}\Gamma(q_i)\bigg(\frac{4}{T_i-T_{i-1}}\bigg)^{q_i-1}. \end{equation*}$

We complete the proof.

Remark 4.3. We notice that if $r = 0$ and $q(t) = q$ , $q$ is a constant, i.e., BVP (1.1) is a fractional differential equation with constant order, then by similar arguments as done in ^[22], we get

$\begin{equation*} \int_{0}^{T}h(s)ds \gt \Gamma(q)\bigg(\frac{4}{T}\bigg)^{q-1}. \end{equation*}$

So, the inequalities (4.2) is a generalized Lyapunov-type inequality for the boundary value problem (1.1).

5. Examples

Example 5.1. Let us consider the following nonlinear boundary value problem

$\begin{equation} \begin{cases} D_{0^+}^{q(t)}x(t)+t^{-0.5}\frac{|x|^{\frac 12}}{1+x^2} = 0, \; \; 0 \lt t \lt 2, \\ u(0) = u(2) = 0, \end{cases} \end{equation}$

(5.1)

where

$q(t) = \begin{cases} 1.2, \ \ 0\leq t\leq 1, \\ 1.6, \ \ 1 \lt t\leq 2. \end{cases}$

We see that $q(t)$ satisfies condition $(H_1)$ ; $t^{0.5}f(t, x(t)) = \frac{|x(t)|^{\frac 12}}{1+x(t)^2}:[0, 2]\times \mathbb{R}\rightarrow \mathbb{R}$ is continuous. Moreover, we have

$t^{0.5}|f(t, x(t))| = \frac{|x(t)|^{\frac 12}}{1+x(t)^2}\leq |x(t)|^{\frac 12}.$

Let $r = 0.5$ , $c_1 = c_2 = 1$ and $\gamma = \frac 12$ . We could verify that $f(t, x) = t^{-0.5}\frac{|x|^{\frac 12}}{1+x^2}$ satisfies condition $(H_2)$ . This suggests that the boundary value problem (5.1) has a solution by the conclusion of Theorem 3.3.

Example 5.2. Let us consider the following linear boundary value problem

$\begin{equation} \begin{cases} D_{0^+}^{q(t)}x(t)+t^{0.4} = 0, \; \; 0 \lt t \lt 3, \\ u(0) = 0, \; u(3) = 0, \end{cases} \end{equation}$

(5.2)

where

$q(t) = \begin{cases} 1.2, \ \ 0\leq t\leq 1, \\ 1.5, \ \ 1 \lt t\leq 2, \\ 1.8, \ \ 2 \lt t\leq 3.\end{cases}$

We see that $q(t)$ satisfies condition $(H_1)$ ; $f(t, x(t)) = t^{0.4}:[0, 3]\times \mathbb{R}\rightarrow \mathbb{R}$ is continuous. Moreover, $|f(t, x(t))| = t^{0.4}\leq 3^{0.4}$ , thus we could take suitable constants to verify $f(t, x) = t^{0.4}$ satisfies condition $(H_2)$ . Then Theorem 3.3 assures the boundary value problem (5.2) has a solution.

In fact, we know that equation of (5.1) can been divided into three expressions as following

$\begin{equation} D_{0+}^{1.2} x(t)+t^{0.4} = 0, \ \ 0 \lt t\leq 1. \end{equation}$

(5.3)

For $1 < t\leq 2$ ,

$\begin{equation} \frac {d^2}{dt^2}\bigg(\int_0^1\frac{(t-s)^{-0.2}} {\Gamma(0.8)}x(s)ds+\int_1^t\frac{(t-s)^{-0.5}} {\Gamma(0.5)}x(s)ds\bigg)+t^{0.4} = 0. \end{equation}$

(5.4)

For $2 < t\leq 3$ ,

$\begin{equation} \frac {d^2}{dt^2} \bigg(\int_0^1\frac{(t-s)^{-0.2}} {\Gamma(0.8)}x(s)ds+\int_1^2\frac{(t-s)^{-0.5}} {\Gamma(0.5)}x(s)ds+\int_2^t\frac{(t-s)^{-0.8}} {\Gamma(0.2)}x(s)ds\bigg)+t^{0.4} = 0. \end{equation}$

(5.5)

By ^[18], we can easily obtain that the following boundary value problems

$\begin{cases} D_{0+}^{1.2} x(t)+t^{0.4} = 0, \ \ 0 \lt t\leq 1, \\ x(0) = 0, x(1) = 0\end{cases}$

$\begin{cases} D_{1+}^{1.5} x(t) = \frac {d^2}{dt^2}\int_1^t\frac{(t-s)^{-0.5}} {\Gamma(0.5)}x(s)ds+t^{0.4} = 0, \ \ 1 \lt t \lt 2, \\ x(1) = 0, x(2) = 0\end{cases}$

$\begin{cases} D_{2+}^{1.8} x(t) = \frac {d^2}{dt^2}\int_2^t\frac{(t-s)^{-0.8}} {\Gamma(0.2)}x(s)ds+t^{0.4} = 0, \ \ 2 \lt t \lt 3, \\ x(2) = 0, x(3) = 0\end{cases}$

respectively have solutions

$\begin{align*} x_1(t)& = \frac{\Gamma(1.4)}{\Gamma(2.6)} \big(t^{0.2}-t^{1.6}\big)\in C[0, 1];\\ \widetilde{x}_2(t)& = \frac{\Gamma(1.4)}{\Gamma(2.9)} ((t-1)^{0.5}-(t-1)^{1.9})\in C[1, 2];\\ \widetilde{x}_3(t)& = \frac{\Gamma(1.4)}{\Gamma(3.2)} ((t-2)^{0.8}-(t-2)^{2.2})\in C[2, 3]. \end{align*}$

It is known by calculation that

$\begin{equation} x_1(t), 0\leq t\leq 1, \quad \quad\quad x_2(t) = \begin{cases} 0, \quad 0\leq t\leq 1, \\ \widetilde{x}_2(t), 1 \lt t\leq 2, \end{cases} x_3(t) = \begin{cases} 0, \quad 0\leq t\leq 2, \\ \widetilde{x}_3(t), 2 \lt t\leq 3, \end{cases} \end{equation}$

(5.6)

are the solutions of (5.3)–(5.5), respectively. By Definition 3.2 and (5.6), we know that

$\begin{equation*} x(t) = \begin{cases} x_1(t) = \frac{\Gamma(1.4)}{\Gamma(2.6)} \big(t^{0.2}-t^{1.6}\big), \quad\quad\quad\quad \quad\quad 0\leq t\leq 1, \\ x_2(t) = \begin{cases} 0, \quad\quad\quad\quad \quad \; \; \quad\quad\quad\quad\quad\quad\quad \quad 0\leq t\leq 1, \\ \frac{\Gamma(1.4)}{\Gamma(2.9)} ((t-1)^{0.5}-(t-1)^{1.9}), \; \; \; 1 \lt t\leq 2, \\ \end{cases} \\ x_3(t) = \begin{cases} 0, \quad\quad \quad \; \; \quad\quad\quad\quad\quad\quad\quad\quad\quad \quad 0\leq t\leq 2, \\ \frac{\Gamma(1.4)}{\Gamma(3.2)} ((t-2)^{0.8}-(t-2)^{2.2}), \; \; \; 2 \lt t\leq 3 \end{cases} \end{cases} \end{equation*}$

is one solution of the boundary value problem (5.2).

6. Conclusions

In this paper, we consider a two-points boundary value problem of differential equations of variable order, which is a piecewise constant function. Based the essential difference about the variable order fractional calculus (derivative and integral) and the integer order and the constant fractional order calculus (derivative and integral), we carry on essential analysis to the boundary value problem (1.1). According to our analysis, we give the definition of solution to the boundary value problem (1.1). The existence result of solution to the boundary value problem (1.1) is derived. We present a Lyapunov-type inequality for the boundary value problems (1.1). Since the variable order fractional calculus (derivative and integral) and the integer order and the constant fractional order calculus (derivative and integral) has the essential difference, it is interesting and challenging about the existence, uniqueness of solutions, Lyapunov-type inequality, etc, to the boundary value problems of differential equations of variable order.

Acknowledgments

This research is supported by the Natural Science Foundation of China (11671181). The authors are thankful to the referees for their careful reading of the manuscript and insightful comments.

Conflict of interest

The author declares no conflicts of interest in this paper.

References

[1]	A. Ahmadi-Javid, Entropic value-at-risk: A new coherent risk measure, J. Optim. Theory Appl., 155 (2012), 1105–1123. https://dx.doi.org/10.1007/s10957-011-9968-2 doi: 10.1007/s10957-011-9968-2
[2]	Y. Amihud, H. Mendelson, Liquidity and stock returns, Financ. Anal. J., 42 (1986), 43–48. http://dx.doi.org/10.2469/faj.v42.n3.43 doi: 10.2469/faj.v42.n3.43
[3]	P. Artzner, F. Delbaen, J.-M. Eber, D. Heath, Thinking coherently, Risk, 10 (1997), 68–71.
[4]	P. Artzner, F. Delbaen, J.-M. Eber, D. Heath, Coherent measures of risk, Math. Financ., 9 (1999), 203–228. http://dx.doi.org/10.1111/1467-9965.00068
[5]	A. Axelrod, L. Carlone, G. Chowdhary, S. Karaman, Data-driven prediction of EVAR with confidence in time-varying datasets, 2016 IEEE 55th Conference on Decision and Control (CDC), Las Vegas, NV, USA, 2016, 5833–5838. http://dx.doi.org/10.1109/CDC.2016.7799166
[6]	D. P. Baron, On the utility theoretic foundations of mean-variance analysis, J. Financ., 32 (1977), 1683–1697. http://dx.doi.org/10.1111/j.1540-6261.1977.tb03363.x doi: 10.1111/j.1540-6261.1977.tb03363.x
[7]	A. Ben-Tal, D. Bertsimas, D. B. Brown, A soft robust model for optimization under ambiguity, Oper. Res., 58 (2010), 1220–1234. http://dx.doi.org/10.1287/opre.1100.0821 doi: 10.1287/opre.1100.0821
[8]	D. Bertsimas, D. B. Brown, Constructing uncertainty sets for robust linear optimization, Oper. Res., 57 (2009), 1483–1495. http://dx.doi.org/10.1287/opre.1080.0646 doi: 10.1287/opre.1080.0646
[9]	M. Borkovec, I. Domowitz, B. Kiernan, V. Serbin, Portfolio optimization and the cost of trading, J. Invest., 19 (2010), 63–76. http://dx.doi.org/10.3905/joi.2010.19.2.063 doi: 10.3905/joi.2010.19.2.063
[10]	J. P. Bouchaud, J. Bonart, J. Donier, M. Gould, Trades, quotes and prices: financial markets under the microscope, Cambridge: Cambridge University Press, 2018. http://dx.doi.org/https://doi.org/10.1017/9781316659335
[11]	S. P. Boyd, L. Vandenberghe, Convex optimization, Cambridge: Cambridge University Press, 2004.
[12]	S. Caçador, J. M. Dias, P. Godinho, Portfolio selection under uncertainty: a new methodology for computing relative-robust solutions, Int. T. Oper. Res., 28 (2021), 1296–1329. http://dx.doi.org/10.1111/itor.12674 doi: 10.1111/itor.12674
[13]	D. Cajas, Entropic portfolio optimization: a disciplined convex programming framework, SSRN Electronic Journal, 2021 (2021), 3792520.
[14]	A. H. Chen, F. J. Fabozzi, D. S. Huang, Portfolio revision under mean-variance and mean-cvar with transaction costs, Rev. Quant. Finan. Acc., 39 (2012), 509–526. http://dx.doi.org/10.1007/s11156-012-0292-1 doi: 10.1007/s11156-012-0292-1
[15]	S. Chennaf, J. B. Amor, Entropic value at risk to find the optimal uncertain random portfolio, Soft Comput., 27 (2023), 15185–15197. http://dx.doi.org/10.1007/s00500-023-08547-5 doi: 10.1007/s00500-023-08547-5
[16]	J. Dufitinema, S. Pynnönen, T. Sottinen, Maximum likelihood estimators from discrete data modeled by mixed fractional brownian motion with application to the nordic stock markets, Commun. Stat. Simul. C., 51 (2022), 5264–5287. http://dx.doi.org/10.1080/03610918.2020.1764581 doi: 10.1080/03610918.2020.1764581
[17]	P. Embrechts, S. I. Resnick, G. Samorodnitsky, Extreme value theory as a risk management tool, N. Am. Actuar. J., 3 (1999), 30–41. http://dx.doi.org/10.1080/10920277.1999.10595797 doi: 10.1080/10920277.1999.10595797
[18]	D. Goldfarb, G. Iyengar, Robust portfolio selection problems, Math. Oper. Res., 28 (2003), 1–38. http://dx.doi.org/10.1287/moor.28.1.1.14260 doi: 10.1287/moor.28.1.1.14260
[19]	F. Hooshmand, Z. Anoushirvani, S. A. MirHassani, Model and efficient algorithm for the portfolio selection problem with real-world constraints under value-at-risk measure, Int. T. Oper. Res., 30 (2023), 2665–2690. http://dx.doi.org/10.1111/itor.13239 doi: 10.1111/itor.13239
[20]	R. P. Huang, Z. S. Xu, S. J. Qu, X. G. Yang, M. Goh, Robust portfolio selection with distributional uncertainty and integer constraints, J. Oper. Res. Soc. China, 11 (2023), 1–27. http://dx.doi.org/10.1007/s40305-023-00466-4 doi: 10.1007/s40305-023-00466-4
[21]	G. Kara, A. Özmen, G.-W. Weber, Stability advances in robust portfolio optimization under parallelepiped uncertainty, Cent. Eur. J. Oper. Res., 27 (2019), 241–261. http://dx.doi.org/10.1007/s10100-017-0508-5 doi: 10.1007/s10100-017-0508-5
[22]	J. Kriens, J. T. van Lieshout, Notes on the Markowitz portfolio selection method, Stat. Neerl., 42 (1988), 181–191. http://dx.doi.org/10.1111/j.1467-9574.1988.tb01232.x doi: 10.1111/j.1467-9574.1988.tb01232.x
[23]	W. Liu, L. Yang, B. Yu, Distributionally robust optimization based on Kernel density estimation and mean-entropic value-at-risk, INFORMS Journal on Optimization, 5 (2022), 68–91. http://dx.doi.org/10.1287/ijoo.2022.0076 doi: 10.1287/ijoo.2022.0076
[24]	G. M. Luo, Mixed complementarity problems for robust optimization equilibrium in bimatrix game, Appl. Math., 57 (2012), 503–520. http://dx.doi.org/10.1007/s10492-012-0029-4 doi: 10.1007/s10492-012-0029-4
[25]	H. Markowitz, Portfolio selection, J. Financ., 7 (1952), 77–91. http://dx.doi.org/10.1111/j.1540-6261.1952.tb01525.x
[26]	E. J. Menvouta, S. Serneels, T. Verdonck, Portfolio optimization using cellwise robust association measures and clustering methods with application to highly volatile markets, Journal of Finance and Data Science, 9 (2023), 100097. http://dx.doi.org/10.1016/j.jfds.2023.100097 doi: 10.1016/j.jfds.2023.100097
[27]	J. E. Mitchell, S. Braun, Rebalancing an investment portfolio in the presence of convex transaction costs, submitted for publication.
[28]	J. E. Mitchell, S. Braun, Rebalancing an investment portfolio in the presence of convex transaction costs, including market impact costs, Optim. Method. Softw., 28 (2013), 523–542. http://dx.doi.org/10.1080/10556788.2012.717940 doi: 10.1080/10556788.2012.717940
[29]	K. Muthuraman, S. Kumar, Multidimensional portfolio optimization with proportional transaction costs, Math. Financ., 16 (2006), 301–335. http://dx.doi.org/10.1111/j.1467-9965.2006.00273.x doi: 10.1111/j.1467-9965.2006.00273.x
[30]	G. C. Pflug, Some remarks on the value-at-risk and the conditional value-at-risk, In: Probabilistic constrained optimization, Boston: Springer, 2000,272–281. http://dx.doi.org/10.1007/978-1-4757-3150-7_15
[31]	R. T. Rockafellar, S. Uryasev, Optimization of conditional value-at-risk, J. Risk, 3 (2000), 21–41.
[32]	R. Sehgal, A. Mehra, Robust reward–risk ratio portfolio optimization, Int. T. Oper. Res., 28 (2021), 2169–2190. http://dx.doi.org/10.1111/itor.12652 doi: 10.1111/itor.12652
[33]	R. L. Sun, T. F. Ma, S. Z. Liu, Portfolio selection based on semivariance and distance correlation under minimum variance framework, Stat. Neerl., 73 (2019), 373–394. http://dx.doi.org/10.1111/stan.12174 doi: 10.1111/stan.12174
[34]	X. J. Tong, F. L. Wu, L. Q. Qi, Worst-case cvar based portfolio optimization models with applications to scenario planning, Optim. Method. Softw., 24 (2009), 933–958. http://dx.doi.org/10.1080/10556780902865942 doi: 10.1080/10556780902865942
[35]	X. J. Tong, F. L. Wu, Robust reward–risk ratio optimization with application in allocation of generation asset, Optimization, 63 (2014), 1761–1779. http://dx.doi.org/10.1080/02331934.2012.672419 doi: 10.1080/02331934.2012.672419
[36]	X. T. Wang, Z. Li, L. Zhuang, Risk preference, option pricing and portfolio hedging with proportional transaction costs, Chaos Soliton. Fract., 95 (2017), 111–130. https://dx.doi.org/10.1016/j.chaos.2016.12.010 doi: 10.1016/j.chaos.2016.12.010
[37]	L. J. Xu, Y. J. Zhou, New robust reward-risk ratio models with CVaR and standard deviation, J. Math., 2022 (2022), 8304411. http://dx.doi.org/10.1155/2022/8304411 doi: 10.1155/2022/8304411
[38]	C. L. Zheng, Y. Chen, Portfolio selection based on relative entropy coherent risk measure, Systems Engineering-Theory & Practice, 34 (2014), 648–655.
[39]	Y. J. Zhou, L. Yang, L. J. Xu, B. Yu, Inseparable robust reward–risk optimization models with distribution uncertainty, Japan J. Indust. Appl. Math., 33 (2016), 767–780. http://dx.doi.org/10.1007/s13160-016-0230-z doi: 10.1007/s13160-016-0230-z
[40]	S. S. Zhu, D. Li, S. Y. Wang, Robust portfolio selection under downside risk measures, Quant. Financ., 9 (2009), 869–885. http://dx.doi.org/10.1080/14697680902852746 doi: 10.1080/14697680902852746
[41]	S. S. Zhu, M. Fukushima, Worst-case conditional value-at-risk with application to robust portfolio management, Oper. Res., 57 (2009), 1155–1168. http://dx.doi.org/10.1287/opre.1080.0684 doi: 10.1287/opre.1080.0684
[42]	J. X. Zhu, Optimal financing and dividend distribution with transaction costs in the case of restricted dividend rates, ASTIN Bull., 47 (2017), 239–268. http://dx.doi.org/10.1017/asb.2016.29 doi: 10.1017/asb.2016.29

This article has been cited by:

1.	Ibtesam Alshammari, Islam M. Taha, On fuzzy soft $\beta$ -continuity and $\beta$ -irresoluteness: some new results, 2024, 9, 2473-6988, 11304, 10.3934/math.2024554
2.	D. I. Taher, R. Abu-Gdairi, M. K. El-Bably, M. A. El-Gayar, Decision-making in diagnosing heart failure problems using basic rough sets, 2024, 9, 2473-6988, 21816, 10.3934/math.20241061
3.	Fahad Alsharari, Ahmed O. M. Abubaker, Islam M. Taha, On $r$ -fuzzy soft $\gamma$ -open sets and fuzzy soft $\gamma$ -continuous functions with some applications, 2025, 10, 2473-6988, 5285, 10.3934/math.2025244
4.	Fahad Alsharari, Hind Y. Saleh, Islam M. Taha, Some Characterizations of k-Fuzzy γ-Open Sets and Fuzzy γ-Continuity with Further Selected Topics, 2025, 17, 2073-8994, 678, 10.3390/sym17050678

Reader Comments

Your name:*

Email:*
© 2024 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)