摘要：
In this article, we define the notion of n -cotorsion pairs in triangulated categories, which is a generalization of the classical cotorsion pairs. We prove that any mutation of an n -cotorsion pair is again an n -cotorsion pair. When n = 1 , this result generalizes the work of Zhou and Zhu for classical cotorsion pairs. As applications, we give a geometric characterization of n -cotorsion pairs in n -cluster categories of type A and give a geometric realization of mutation of n -cotorsion pairs via rotation of certain configurations of n -diagonals.
In this article, we define the notion of n -cotorsion pairs in triangulated categories, which is a generalization of the classical cotorsion pairs. We prove that any mutation of an n -cotorsion pair is again an n -cotorsion pair. When n = 1 , this result generalizes the work of Zhou and Zhu for classical cotorsion pairs. As applications, we give a geometric characterization of n -cotorsion pairs in n -cluster categories of type A and give a geometric realization of mutation of n -cotorsion pairs via rotation of certain configurations of n -diagonals.
期刊：
International Journal of Forecasting,2025年 ISSN：0169-2070
通讯作者：
Jie Kang
作者机构：
[Zhifeng Dai] School of Mathematics and Statistics, Changsha University of Science and Technology, Changsha, China;[Fuwei Jiang] Center for Macroeconomic Research, School of Economics and Wang Yanan Institute for Studies in Economics, Xiamen University, Xiamen, China;[Jie Kang] China Economics and Management Academy, Central University of Finance and Economics, Beijing, China;[Bowen Xue] School of Economics and Wang Yanan Institute for Studies in Economics, Xiamen University, Xiamen, China
通讯机构：
[Jie Kang] C;China Economics and Management Academy, Central University of Finance and Economics, Beijing, China
摘要：
This paper investigates the role of time–frequency information in dimension reduction prediction of stock returns. Using the long-term wavelet component of monthly S&P500 excess returns as supervision, we employ a machine learning method to extract the common predictive factor from prevalent macroeconomic variables and construct a new macroeconomic index aligned with stock return prediction. The macroeconomic index exhibits significant predictive power, both in and out of sample, at the market and portfolio levels. It outperforms all individual macroeconomic predictors and the factors based on higher frequency information of realized returns. Our findings demonstrate substantial economic value of the new index in asset allocation. Moreover, we also observe a complementary relation between macroeconomic index and investor sentiment. The predictive power is most pronounced during high-economic-uncertainty periods when investors are likely to underreact to fundamental signals and stems from cash flow predictability channel.
This paper investigates the role of time–frequency information in dimension reduction prediction of stock returns. Using the long-term wavelet component of monthly S&P500 excess returns as supervision, we employ a machine learning method to extract the common predictive factor from prevalent macroeconomic variables and construct a new macroeconomic index aligned with stock return prediction. The macroeconomic index exhibits significant predictive power, both in and out of sample, at the market and portfolio levels. It outperforms all individual macroeconomic predictors and the factors based on higher frequency information of realized returns. Our findings demonstrate substantial economic value of the new index in asset allocation. Moreover, we also observe a complementary relation between macroeconomic index and investor sentiment. The predictive power is most pronounced during high-economic-uncertainty periods when investors are likely to underreact to fundamental signals and stems from cash flow predictability channel.
期刊：
JOURNAL OF THE LONDON MATHEMATICAL SOCIETY-SECOND SERIES,2025年111(2):e70091- ISSN：0024-6107
通讯作者：
Xu, RZ
作者机构：
[Lin, Qiang] Changsha Univ Sci & Technol, Sch Math & Stat, Changsha, Peoples R China.;[Xu, Runzhang; Lin, Qiang] Harbin Engn Univ, Coll Math Sci, Harbin 150001, Peoples R China.
通讯机构：
[Xu, RZ ] H;Harbin Engn Univ, Coll Math Sci, Harbin 150001, Peoples R China.
摘要：
In this paper, we conduct a comprehensive study of the global well-posedness of solution for a class of nonlocal wave equations with variable-order fractional Laplacian and variable exponent nonlinearity by constructing a suitable framework of the variational theory. We first prove the local-in-time existence of the weak solution via the Galerkin approximation technique and fixed point theory. Then by constructing the potential well theory, we classify the initial data leading to the global existence and finite time blowup of the solution for three different initial energy cases, that is, subcritical initial energy case, critical initial energy case, and supercritical initial energy case. For the subcritical and critical initial energy cases, we show that the solution exists globally in time when the initial data belong to the stable manifold and blows up in finite time when the initial data belong to the unstable manifold. For the supercritical initial energy case, we observe some initial conditions that enable the finite time blow-up solution by an adapted concavity method, and the issue of global existence still remains unsolved. As a further study of finite time blowup, we estimate the upper and lower bounds of blow-up time by using different strategies, that is, applying some first-order differential inequality regardless of the different initial energy levels, to give a unified expression for the lower bound estimation for three initial energy levels. For the upper bound estimation, we utilize two second-order differential inequalities influenced by the different energy levels to give the upper bound estimations of the blow-up time at each initial energy level.
摘要：
This paper focuses the initial boundary value problem to the compressible Oldroyd-B model in the two-dimensional space with vacuum. We establish a new blowup criterion in terms of parallel to rho parallel to L infinity tBMOx only. This work can be viewed as an extension of Lu and Zhang [Relative entropy, weak-strong uniqueness, and conditional regularity for a compressible Oldroyd-B model, SIAM J. Math. Anal., 50(2018), 557-590] where a blowup criterion involving the upper bound of the fluid density in the two-dimensional setting has been given.
期刊：
ANNALI DI MATEMATICA PURA ED APPLICATA,2025年:1-32 ISSN：0373-3114
通讯作者：
Yu, MZ
作者机构：
[Xie, Weihong] Cent South Univ, Sch Math & Stat, HNP, LAMA, Changsha 410083, Hunan, Peoples R China.;[Yu, Mingzhu; Yu, MZ] Changsha Univ Sci & Technol, Sch Math & Stat, Changsha 410083, Hunan, Peoples R China.
通讯机构：
[Yu, MZ ] C;Changsha Univ Sci & Technol, Sch Math & Stat, Changsha 410083, Hunan, Peoples R China.
关键词：
FitzHugh-Nagumo system;Infinitely many positive solutions;Sign-changing solutions;Gluing methods
摘要：
In this article we present the existence of infinitely many non-radial positive or sign-changing solutions for the following FitzHugh–Nagumosystem:
$$\begin{aligned} \left\{ \begin{array}{ll} \Delta u-a(|x|)u+g(u)-\delta v=0, \quad & x\in \mathbb {R}^N,\\ \Delta v+u=0, & x\in \mathbb {R}^N,\\ u(x), ~v(x)\rightarrow 0, & \text{ as }~ |x|\rightarrow +\infty ,\\ \end{array}\right. \end{aligned}$$
where
$$N\ge 5$$
,
$$\delta >0$$
,
$$g(u)=(a_0+1)u^2-u^3$$
,
$$0<a_0<\frac{1}{2}$$
and
$$a(|x|)\in (0,\frac{1}{2})$$
satisfies some decay conditions at the infinity. More precisely, for any positive integer k large, there is a
$$\delta _k>0$$
such that for
$$0<\delta <\delta _k$$
, there exists positive solutions with 2k peaks, which are respectively concentrated at the vertices of a regular k-polygon on two circles in 3-dimensional space with the radium
$$r\sim k \ln k$$
and the height
$$h\sim \frac{1}{k}$$
. In addition, the sign-changing solutions with 2k peaks are evenly distributed on the equatorial
$$\textrm{T}=\{x\in \mathbb {R}^2:x_1^2+x_2^2=r^2\}$$
in the
$$(x_1, x_2)$$
-plane. As a by-product, we give the similar results of Schödinger-Poisson in
$$\mathbb {R}^N$$
for
$$N\ge 3$$
.
期刊：
Nonlinear Analysis: Real World Applications,2025年82:104226 ISSN：1468-1218
通讯作者：
Chuangxia Huang
作者机构：
College of Science, Hunan University of Science and Engineering, Yongzhou 425199, Hunan, China;[Xiaodan Ding] School of Mathematics and Statistics, Changsha University of Science and Technology;Hunan Provincial Key Laboratory of Mathematical Modeling and Analysis in Engineering, Changsha, 410114, Hunan, China;[Chuangxia Huang] College of Science, Hunan University of Science and Engineering, Yongzhou 425199, Hunan, China<&wdkj&>School of Mathematics and Statistics, Changsha University of Science and Technology;[Chuangxia Huang; Xiaodan Ding] Hunan Provincial Key Laboratory of Mathematical Modeling and Analysis in Engineering, Changsha, 410114, Hunan, China
通讯机构：
[Chuangxia Huang] C;College of Science, Hunan University of Science and Engineering, Yongzhou 425199, Hunan, China<&wdkj&>School of Mathematics and Statistics, Changsha University of Science and Technology;Hunan Provincial Key Laboratory of Mathematical Modeling and Analysis in Engineering, Changsha, 410114, Hunan, China
摘要：
The precise baseline determination (PBD) is a crucial technique for spaceborne Interferometric Synthetic Aperture Radar (InSAR) formation-flying satellites to successfully carry out scientific missions. The Chinese InSAR formation T-A and T-B satellites are equipped with the multi-GNSS receiver that can track Global Positioning System (GPS) and BeiDou satellite navigation System (BDS-3) signals for PBD. First, we evaluate the data quality of spaceborne GPS and BDS-3 observations and carry out the absolute precise orbit determination (POD) with single-satellite ambiguity fixed, respectively. The result shows that a three-dimensional (3D) POD consistency of 1.48 cm (RMS) between them is reached. Then, the baseline products are obtained based on GPS and BDS-3 observations, respectively. The PBD of GPS-based is assessed via internal consistency checks and external baseline validations, Internally, the overlap comparison is sub-millimeter. Externally, comparisons with external agency is 1.36 mm in 3D. Taking the baseline results based on GPS as a reference, the difference shows that a consistency of PBD result based on BDS-3 is 0.91 mm in 3D. Finally, we perform the GPS + BDS-3 multi-GNSS data fusion for PBD experiment. The result shows that combining multi-GNSS data can not only increase the position precision and obtain a higher overlapping accuracy, but also improve the internal consistency of baseline products.
The precise baseline determination (PBD) is a crucial technique for spaceborne Interferometric Synthetic Aperture Radar (InSAR) formation-flying satellites to successfully carry out scientific missions. The Chinese InSAR formation T-A and T-B satellites are equipped with the multi-GNSS receiver that can track Global Positioning System (GPS) and BeiDou satellite navigation System (BDS-3) signals for PBD. First, we evaluate the data quality of spaceborne GPS and BDS-3 observations and carry out the absolute precise orbit determination (POD) with single-satellite ambiguity fixed, respectively. The result shows that a three-dimensional (3D) POD consistency of 1.48 cm (RMS) between them is reached. Then, the baseline products are obtained based on GPS and BDS-3 observations, respectively. The PBD of GPS-based is assessed via internal consistency checks and external baseline validations, Internally, the overlap comparison is sub-millimeter. Externally, comparisons with external agency is 1.36 mm in 3D. Taking the baseline results based on GPS as a reference, the difference shows that a consistency of PBD result based on BDS-3 is 0.91 mm in 3D. Finally, we perform the GPS + BDS-3 multi-GNSS data fusion for PBD experiment. The result shows that combining multi-GNSS data can not only increase the position precision and obtain a higher overlapping accuracy, but also improve the internal consistency of baseline products.
期刊：
JOURNAL OF SUPERCOMPUTING,2025年81(3):1-22 ISSN：0920-8542
通讯作者：
Hongfang Gong
作者机构：
[Minyi Ma; Hongfang Gong; Yingjing Ding] School of Mathematics and Statistics, Changsha University of Science and Technology, Changsha, China
通讯机构：
[Hongfang Gong] S;School of Mathematics and Statistics, Changsha University of Science and Technology, Changsha, China
关键词：
Text classification;Graph neural networks;Transformer;Multi-head attention
摘要：
Text classification is a foundational natural language processing task. Many models have transformed text data into innovative graph structures and employed graph neural networks (GNNs) to learn representations for classification. However, graphs constructed based on artificial rules may contain redundant connections, which can introduce noise. Additionally, graph neural network-based (GNN-based) models are ineffective in learning word order. To cope with these difficulties, we propose graph affine Transformer with a symmetric adaptation strategy (GATSAS) for text classification. After graph construction, a symmetric gated graph neural network is crafted to detect edge attributes and remove trivial connections. In the graph learning step, graph affine Transformer encodes word positions and updates word representations with multi-head graph affine attention. Then, the readout module fuses word features to achieve text representation. The experimental results show that GATSAS has an average accuracy of 0.11
$$-$$
2.37% higher than the GNN-based benchmark models.
期刊：
Mathematics and Computers in Simulation,2025年233:389-399 ISSN：0378-4754
通讯作者：
Li, JL
作者机构：
[Kuang, Daipeng] Changsha Univ Sci & Technol, Sch Math & Stat, Changsha 410114, Hunan Province, Peoples R China.;[Kuang, Daipeng; Li, Jianli; Li, JL] Hunan Normal Univ, Sch Math & Stat, Changsha 410081, Hunan Province, Peoples R China.;[Gao, Dongdong] Tongling Univ, Dept Math & Comp Sci, Tongling 244000, Anhui Province, Peoples R China.
通讯机构：
[Li, JL ] H;Hunan Normal Univ, Sch Math & Stat, Changsha 410081, Hunan Province, Peoples R China.
关键词：
Asymptotic stability;Event-triggered impulsive control (ETIC);Finite-time stability;Zeno behavior;Delay-dependent impulses
摘要：
This paper addresses the problem of stabilizing nonlinear stochastic systems using an event-triggered impulse mechanism (ETIM) and control theory. Sufficient criteria for achieving asymptotic stability (AS), finite-time stability (FTS), and finite-time contraction stability (FTCS) are obtained. In the ETIM, respectively, the timer threshold and free-control indexes are introduced to effectively prevent Zeno behavior and unnecessary impulses, thus conserving control resources. Furthermore, the impulse policy is formulated by considering both the current state and past information of the system, resulting in the generation of impulses that encompass both common and delay-dependent characteristics. The research findings reveal that the stability of the system is influenced by the stochastic system, impulse strength, time delay, and the ETIM. It is demonstrated that the impulse strength and the impulse sequence are the primary factors contributing to system stability, while time delay in impulse has a negative impact. The obtained criteria are applied to a stochastic network system, and the validity of the results is supported through illustrative examples and numerical simulations.
This paper addresses the problem of stabilizing nonlinear stochastic systems using an event-triggered impulse mechanism (ETIM) and control theory. Sufficient criteria for achieving asymptotic stability (AS), finite-time stability (FTS), and finite-time contraction stability (FTCS) are obtained. In the ETIM, respectively, the timer threshold and free-control indexes are introduced to effectively prevent Zeno behavior and unnecessary impulses, thus conserving control resources. Furthermore, the impulse policy is formulated by considering both the current state and past information of the system, resulting in the generation of impulses that encompass both common and delay-dependent characteristics. The research findings reveal that the stability of the system is influenced by the stochastic system, impulse strength, time delay, and the ETIM. It is demonstrated that the impulse strength and the impulse sequence are the primary factors contributing to system stability, while time delay in impulse has a negative impact. The obtained criteria are applied to a stochastic network system, and the validity of the results is supported through illustrative examples and numerical simulations.
摘要：
We study the long-term behavior of the non -autonomous wave equations with a new -type nonlocal weak damping and critical nonlinearity. After proving the global well-posedness, we first study the asymptotic regularity of solutions, which shows that the solutions are exponentially approaching a more regular bounded subset. Based on this regularity result, we finally obtain the existence and smoothness of uniform exponential attractors.
摘要：
Uncertainty relations are a distinctive characteristic of quantum physics. We provide a set of uncertainty relations for quantum channels based on the generalized Wigner-Yanase skew information. Both product-form and summation-form uncertainty inequalities are derived for two general quantum channels, which include the ones given by the Wigner-Yanase skew information as special cases. These uncertainty inequalities are further refined as uncertainty sequences. Detailed examples are given for several prototypical quantum channels.
摘要：
The bilinear model has been frequently used in fields like control theory and economics to model seismic data. In this article, time-functional variance (TFV) noises are embedded into a specific bilinear model. We propose a generalized autoregressive conditional heteroskedasticity-type maximum likelihood estimator (GMLE) with a sieve method and then provide various inferential techniques based on this GMLE. It is shown that under the finite fourth moment of errors, the GMLE is consistent and asymptotically normally distributed. A simulation study and analysis of real data are additionally carried out to evaluate GMLE’s finite sample performance.
作者机构：
[Wang, Wentao] Shanghai Univ Engn Sci, Sch Math Phys & Stat, Shanghai 201620, Peoples R China.;[Zeng, Wei] Changsha Univ Sci & Technol, Sch Math & Stat, Changsha 410114, Hunan, Peoples R China.;[Chen, Wei; Chen, W] Shanghai Lixin Univ Accounting & Finance, Sch Stat & Math, Shanghai 201209, Peoples R China.
通讯机构：
[Chen, W ] S;Shanghai Lixin Univ Accounting & Finance, Sch Stat & Math, Shanghai 201209, Peoples R China.
关键词：
Inertial neural networks;Global exponential stability;Delay;Characteristics method;Reduced order method;Non-reduced order method
摘要：
In this paper, we utilize the characteristics method to deduce several novel sufficient conditions for the global exponential stability of delayed inertial neural networks (DINNs). The proposed criteria are presented as a set of linear scalar inequalities, which notably do not entail the use of either reduced order method or Lyapunov-Krasovskii functionals (LKFs), distinguishing them from existing results and offering straightforward solvability. Lastly, we substantiate the analytical outcomes through three numerical examples accompanied by respective simulations.
In this paper, we utilize the characteristics method to deduce several novel sufficient conditions for the global exponential stability of delayed inertial neural networks (DINNs). The proposed criteria are presented as a set of linear scalar inequalities, which notably do not entail the use of either reduced order method or Lyapunov-Krasovskii functionals (LKFs), distinguishing them from existing results and offering straightforward solvability. Lastly, we substantiate the analytical outcomes through three numerical examples accompanied by respective simulations.
作者机构：
College of Science, Hunan University of Science and Engineering, Yongzhou, China;[Yanchen Deng; Yaqian Cai; Xian Zhao] School of Mathematics and Statistics, Changsha University of Science and Technology, Changsha, China;[Xiaoguang Yang] Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing, China;[Chuangxia Huang] College of Science, Hunan University of Science and Engineering, Yongzhou, China<&wdkj&>School of Mathematics and Statistics, Changsha University of Science and Technology, Changsha, China
通讯机构：
[Chuangxia Huang] C;College of Science, Hunan University of Science and Engineering, Yongzhou, China<&wdkj&>School of Mathematics and Statistics, Changsha University of Science and Technology, Changsha, China
摘要：
Using a sample of Chinese listed firms for the period 2006–2021, this paper constructs dynamic stock networks annually using symbolization and mutual information methods, and investigates the impact of stock network centrality on one-year-ahead stock price crash risk with the help of the bad news hoarding mechanism. The authors find robust evidence that firms with higher centrality are less likely to experience stock price crashes in the future. An examination of underlying mechanisms suggests that being at the center of the stock network enhances firms' investment efficiency and managers' cost of engaging in earnings management, thereby reducing the likelihood of such firms forming and hoarding bad news, and hence their crash risk. Further analysis reveals that the mitigating effect of network centrality on stock price crashes is more salient for firms with weaker external monitoring and less conservative accounting policies. Overall, this paper sheds light on a novel benefit of being at the center of the stock network, namely that central firms are less prone to crash risk, which provides practical insights for risk-management applications related to asset pricing and tail events.
作者机构：
[Li, Zhen] School of Mathematics and Statistics, Changsha University of Science and Technology, Changsha, Hunan, 410114, China
摘要：
The paper proposed a second-order steady-state-preserving nonstaggered central scheme for solving one-layer and two-layer open channel flows via the flux globalization. The global flux transforms the model into the homogeneous form, avoiding the complex discretization of the source terms. However, when the traditional appropriate quadrature rule discrete the global variables, the scheme tends to maintain only the moving-water equilibrium but not the “lake at rest” equilibrium. This paper proposes a new discretization method, the steady-state discretization (SSD) method of global variables, so that not only the still-water equilibrium can be maintained, but also the moving-water equilibrium, i.e., the discharge, the energy and the global flux are equilibrium. The scheme also ensures that the cross-sectional wet area is positive by introducing a “draining” time-step technique. Numerical experiments verify that the scheme is well-balanced, positivity-preserving and robust when flowing through open channel flows under the continuous or discontinuous bottom topography and channel width, and exactly capturing small perturbations and propagating interfaces of the steady-state solution.
The paper proposed a second-order steady-state-preserving nonstaggered central scheme for solving one-layer and two-layer open channel flows via the flux globalization. The global flux transforms the model into the homogeneous form, avoiding the complex discretization of the source terms. However, when the traditional appropriate quadrature rule discrete the global variables, the scheme tends to maintain only the moving-water equilibrium but not the “lake at rest” equilibrium. This paper proposes a new discretization method, the steady-state discretization (SSD) method of global variables, so that not only the still-water equilibrium can be maintained, but also the moving-water equilibrium, i.e., the discharge, the energy and the global flux are equilibrium. The scheme also ensures that the cross-sectional wet area is positive by introducing a “draining” time-step technique. Numerical experiments verify that the scheme is well-balanced, positivity-preserving and robust when flowing through open channel flows under the continuous or discontinuous bottom topography and channel width, and exactly capturing small perturbations and propagating interfaces of the steady-state solution.
摘要：
Determining precise multi-satellite orbits and geodetic parameters using satellite laser ranging (SLR) observations of spherical laser geodetic satellites is essential for high-accuracy geodetic scientific applications. Weekly multi-satellite combined solutions for satellite orbits, station coordinates and earth rotation parameters (ERPs) using SLR observations from LAGEOS, Etalon and newly launched LARES-2 satellites are achieved using a self-developed software function based on a new processing strategy. We present an in-depth accuracy analysis of the satellite orbits and geodetic parameters determined from the combined solution of LAGEOS-Etalon and LAGEOS-Etalon-LARES-2, in which the impact of two years of LARES-2 SLR observations is mainly studied. The root mean square (RMS) value of post-fit residuals is 0.81 cm for LARES-2, which is comparable to or slightly smaller than that of the LAGEOS and Etalon satellites. The statistics of the differences in overlapping orbit and external orbit comparison are performed for orbit accuracy assessment. For LAGEOS and Etalon, the results of overlapping orbit and external orbit comparison present similar conclusions. The mean of the RMS values of orbit difference for LAGEOS is about 3 cm in the three-dimensional (3D) direction, and less than 1 cm in the radial ( R ) direction. The mean RMS value of orbit difference for Etalon is about 15 cm in the 3D direction, and less than 3 cm in the R direction. The inclusion of LARES-2 SLR observations can slightly improve the orbit accuracy of Etalon. For LARES-2, the mean RMS values of overlapping orbit difference are 1.79 cm and 0.29 cm in the 3D and R directions, respectively, which are smaller than those for LAGEOS. Moreover, the stability and accuracy of the estimated station coordinates and ERPs are improved after adding LARES-2 SLR observations to multi-satellite combined solutions.
摘要：
In this paper, we first substantiate the global existence of bounded positive solutions for a kind of shunting inhibitory cellular neural networks with D operator and proportional delay. Second, by utilising differential inequality techniques and the Lyapunov functional method, some testable criteria are achieved to verify the global exponential stability of the positive almost periodic solutions for the proposed networks, which encompasses existing ones in the literature as some special cases. Finally, an explanatory example is performed to justify our analytical findings.
摘要：
One of the fundamental issues in predicting firm performance from the perspective of complex systems is how to accurately construct stock networks. Most stock network-based research is mainly limited to traditional econometric models, which suffer from being pairwise, linear, or low-dimensionality. Literature dealing with this issue from the perspective of machine learning seems to be scarce; such investigations are, however, particularly relevant for corporate governance. Using a sample of listed firms in the Chinese A-share market from 2006 to 2021, this paper constructs directed-weighted networks by employing the LightGBM model with genetic algorithm-optimized parameters and then investigates whether stock network centrality is instrumental in predicting firm performance. The empirical results suggest that: (i) the LightGBM model outperforms other GBDT-based techniques in efficiency and prediction accuracy; (ii) the stock network inferred from the GA-LightGBM model serves as an effective tool to reflect market volatility, surpassing classical networks constructed by Granger causality test, error correction model, and elastic network; (iii) firms with higher centrality are more likely to experience inferior performance in the future; (iv) network centrality impairs firm performance through increased financial distress and information opacity, with the financial distress channel playing a larger role.
One of the fundamental issues in predicting firm performance from the perspective of complex systems is how to accurately construct stock networks. Most stock network-based research is mainly limited to traditional econometric models, which suffer from being pairwise, linear, or low-dimensionality. Literature dealing with this issue from the perspective of machine learning seems to be scarce; such investigations are, however, particularly relevant for corporate governance. Using a sample of listed firms in the Chinese A-share market from 2006 to 2021, this paper constructs directed-weighted networks by employing the LightGBM model with genetic algorithm-optimized parameters and then investigates whether stock network centrality is instrumental in predicting firm performance. The empirical results suggest that: (i) the LightGBM model outperforms other GBDT-based techniques in efficiency and prediction accuracy; (ii) the stock network inferred from the GA-LightGBM model serves as an effective tool to reflect market volatility, surpassing classical networks constructed by Granger causality test, error correction model, and elastic network; (iii) firms with higher centrality are more likely to experience inferior performance in the future; (iv) network centrality impairs firm performance through increased financial distress and information opacity, with the financial distress channel playing a larger role.
期刊：
Advances in Continuous and Discrete Models,2025年2025(1):1-12 ISSN：2731-4235
通讯作者：
Lian Duan
作者机构：
[Qian Wang] School of Mathematics and Statistics, Changsha University of Science and Technology, Changsha, P.R. China;[Nan Sun] School of Mathematics and Big Data, Anhui University of Science and Technology, Huainan, P.R. China;Anhui Province Engineering Laboratory for Big Data Analysis and Early Warning Technology of Coal Mine Safety, Huainan, P.R. China;[Lian Duan] School of Mathematics and Big Data, Anhui University of Science and Technology, Huainan, P.R. China<&wdkj&>Anhui Province Engineering Laboratory for Big Data Analysis and Early Warning Technology of Coal Mine Safety, Huainan, P.R. China
通讯机构：
[Lian Duan] S;School of Mathematics and Big Data, Anhui University of Science and Technology, Huainan, P.R. China<&wdkj&>Anhui Province Engineering Laboratory for Big Data Analysis and Early Warning Technology of Coal Mine Safety, Huainan, P.R. China
摘要：
This paper proposes a periodic hematopoiesis model with a nonlinear density-dependent mortality term from the perspective of biological modeling. Employing rigorous analytical skills and Lyapunov functional approach, some new criteria are set up for the existence and global exponential stability of positive periodic solutions for the considered model. Finally, a numerical example is provided to confirm the rationality of the theoretical results. The acquired main results complement and enrich the conclusions in the existing literature.
摘要：
Let C\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\cal C}$$\end{document} be a triangulated category. We define m-term subcategories on C\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\cal C}$$\end{document} induced by n-rigid subcategories, which are extriangulated subcategories of C\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\cal C}$$\end{document}. Then we give a one-to-one correspondence between cotorsion pairs on 2-term subcategories G\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\cal G}$$\end{document} and support tau-tilting subcategories on an abelian quotient of G\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\cal G}$$\end{document}. If an m-term subcategory is induced by a co-t-structure, then we have a one-to-one correspondence between cotorsion pairs on it and cotorsion pairs on C\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\cal C}$$\end{document} under certain conditions.
