Government social assistance is a basic policy tool to prevent and resolve the poverty risk of low-income groups, and it is also a solid base for promoting common prosperity. This paper constructs a common prosperity index based on the data of three Chinese household finance surveys in 2015, 2017 and 2019, and uses the fuzzy breakpoint regression method to empirically test the impact of government social assistance on the common prosperity index. The study found that government social assistance is conducive to improving the family common wealth index, basic living assistance plays a decisive role, other assistance is not obvious, the conclusion is still valid after a series of robustness tests. The influence mechanism of government social assistance on common prosperity has two sides. On the one hand, government social assistance promotes common prosperity of families by expanding income channels, promoting policy coordination and driving consumption upgrading. On the other hand, government social assistance inhibits household common prosperity by inducing welfare dependency and reducing total assets. The results of heterogeneity analysis show that the influence of government social assistance on common prosperity is on the rise over time, and the influence of high Engel coefficient and high dependency ratio and household registration in rural and central and western areas on common prosperity is more significant. The research findings of this paper provide micro-evidence for evaluating the policy effect of government social assistance, and also provide an important reference for exploring the path to achieve common prosperity.
The forecasting of age-specific mortality rates of the national population is the basis for population forecasting, pension cost assessment and longevity risk measurement. However, mortality rate data in China mainland are both limited and subject to significant fluctuations. Traditional mortality rate models exhibit certain deficiencies in making accurate forecasts. Therefore, developing a dynamic forecasting model tailored to the specific characteristics of mortality rate data in China mainland is of paramount importance. Based on this premise, this paper introduces a novel mortality forecasting model based on transfer learning. The model aims to transfer information from the mortality rates of multiple other countries or regions to learn the underlying trends in mortality rate changes in China mainland, thereby improving model performance. Determining suitable population mortality rate datasets for transfer learning is a major challenge. The appropriateness of the selection directly impacts the model’s effectiveness. To address this, the paper proposes a data-driven selection algorithm to efficiently identify suitable datasets. Our results indicate that transfer learning models effectively enhance prediction performance by transferring information from suitable datasets, thereby compensating for the deficiencies of traditional models in predicting population mortality rates in China mainland. Based on the transfer learning model, this paper provides forecasts of remaining life expectancy, offering more accurate data support for pension actuarial assessment and longevity risk measurement.
Against the backdrop of accelerating global climate change, shocks from extreme temperatures have increasingly exacerbated the energy system’s “trilemma”. This paper introduces an improved coupling coordination degree model, utilizing energy trilemma synergy degree as a proxy variable to capture the dynamic trade-offs within the energy system. Based on panel data from 30 provinces in China’s mainland spanning 2004 to2022, our findings reveal that extreme temperatures significantly aggravate the energy trilemma, manifested as a decline in the synergistic coordination of the energy “impossible triangle”. This conclusion remains robust across a series of tests, including the use of the meteorological Niño index as an instrumental variable. A mechanism analysis indicates that extreme temperatures systematically disrupt the equilibrium between energy security, energy equity, and environmental sustainability through three primary channels: Price transmission, fiscal response, and resource utilization. This adverse impact is particularly pronounced in regions characterized by a scarcity of traditional energy endowments, a high concentration of manufacturing, and limited carbon sink capacity. Notably, institutional innovations such as water rights trading markets can effectively counteract this adverse impact, demonstrating strong climate adaptability. Furthermore, spatial econometric modeling reveals that a localized extreme temperature shock not only worsens a region’s own energy predicament but also aggravates the energy system imbalances in neighboring areas through significant negative spatial spillover effects, with the indirect effects outweighing the direct effects. This study contributes a novel analytical framework and new empirical evidence for understanding the nexus between climate physical risks and energy economics. It underscores that establishing a regionally coordinated framework for climate resilience is paramount to achieving a sustainable energy transition.
China’s urban structure has evolved from a single central city to a multi-core urban agglomeration region. It is of great significance to explore differentiated environmental regulation policies for the governance of urban agglomerations. We select 13486 air pollution control policies issued by governments from 2000 to 2018, and extract 32 effective policy themes through the LDA model of the machine learning algorithms. We divide them into administration-oriented, market-oriented, and comprehensive regulation policies, and further explore the difference among different regulation policies on the improvement of air quality in 124 cities and ten urban agglomerations(such as Beijing-Tianjin-Hebei urban agglomeration, Chengdu-Chongqing urban agglomeration, Fenwei plain urban agglomeration, etc.). The results indicate that: 1) Compared with the comprehensive regulation policies, the administration-oriented and the market-oriented policy significantly reduce air pollution emissions. Moreover, the market-oriented environmental regulation policies achive a better effect. 2) Considering the heterogeneity of different urban agglomerations, administration-oriented policies significantly reduces the air pollution in the Yangtze River Delta and the Pearl River Delta. For the Central Plains urban agglomeration, market-oriented incentives exert a more effective role in air pollution in Chengdu-Chongqing and Liaoning region. 3) The total effect of the policy is mainly dominated by the administration-oriented policy, but the impact of the administration-oriented policy on air pollution is gradually saturated. However, there is a great policy space for market-oriented environmental regulation. Thus, government should focus on designing market-oriented regulations when setting environmental regulation policies, and increase the implementation space and strength of market-oriented policies; for region with developed economy and high population density, it is necessary to take administration-oriented environmental regulations, while for the heavy industrial urban agglomerations, policymakers should further strengthen the market-oriented regulations such as fiscal taxation, emissions trading and emission permit trading system, etc.
In the context of China’s comprehensive green transformation, the fundamental way for enterprises to reduce pollution and carbon emissions is technological innovation, and the integration of digital-green technologies is becoming the key to breaking through traditional technological paths. This paper constructs a firm-level index of digital-green technology fusion based on the patent co-classification method and empirically examines its emission abatement effects and mechanisms using data from Chinese A-share listed companies from 2003 to 2021. The results show that digital-green technology fusion significantly reduces PM2.5 concentrations within a1-kilometer radius around firms, meaning that the implementation of technological integration by enterprises can effectively improve the air quality in the surrounding areas. Mechanism analysis indicates that it promotes emission reduction internally by enhancing corporate productivity and broadening green technology adoption, and externally by improving the quality of environmental information disclosure. Heterogeneity analysis indicates that the emission reduction effects are more pronounced in high-tech and heavily polluting industries, highly innovative firms, and firms in the growth stage. Moreover, improved digital infrastructure and intellectual property protection system provide stronger support for digital-green technology integration to exert its emission reduction effect. Further analysis demonstrates that the effect remains significant within a 5
The study of green industrial and financial policies in the context of “dual-carbon” is of great significance for promoting low-carbon transitions and green development. Based on the framework of two-sector economic model, we analyse the mechanism by which green financial policy and green industrial policy can interact to reduce the cost of financing in the green sector and adjust capital structures. The analysis found that green industrial policy, by acting on the demand side, can lead to a reduction in the share of assets in brown sectors and an increase in the share of assets in green sectors. However, implementing green industrial policies alone leads to a suboptimal outcome with reduced financing costs for brown sectors and increased financing costs for green sectors, inducing a rebound effect. To deepen the effect of green transformation, the government needs to implement green financial policies from the supply side, which can reduce the share of brown capital and increase the share of green capital. Implementing green financial policies also brings about increased financing costs in brown sectors and decreased financing costs in green sectors, which directs capital flows away from brown to green sectors. We empirically examine the effects of green industrial and green finance policy mix on capital structure and financing costs. This paper provides the theoretical basis and empirical evidences for green industry and finance policy mix to direct capital towards low-carbon industries for an efficient low-carbon transition.
Accurately quantifying the environmental costs of Chinese provinces and industries participating in global value chains, and effectively reducing the embodied carbon intensity associated with this participation, is crucial for China’s targeted carbon reduction efforts and the promotion of high-quality economic development. Based on the input-output method, this paper separately measures the level of industrial digitalization and the embodied carbon intensity of global value chain participation at the provincial-industry level in China. On this basis, the paper empirically examines the impact and mechanisms of industrial digitalization on global value chain embodied carbon intensity, and further explores the moderating effects of global value chain participation and foreign direct investment. This study finds significant heterogeneity in global value chain embodied carbon intensity and industrial digitalization levels across different provinces and industries. Industrial digitalization significantly reduces the embodied carbon intensity of China’s global value chain participation, primarily through three pathways:Facilitating the green transformation of value chains, optimizing the energy consumption structure, and improving energy efficiency. Moreover, the impact of industrial digitalization in reducing global value chain embodied carbon intensity is more pronounced in the service sector and through the input of digital elements from other provinces. Industrial digitalization plays a more significant role in reducing the embodied carbon intensity of simple global value chain participation and intra-provincial value chains participation. The effect is also stronger in provinces with free trade zones and in energy-intensive industries. Furthermore, industrial digitalization is crucial for reducing the embodied carbon intensity of China’s deep backward global value chain participation, while the inflow of low-quality foreign direct investment hampers the carbon reduction effect of industrial digitalization. These findings provide important policy implications for further promoting the green transformation of global value chains and advancing high-level opening-up.
The pursuit of “dual carbon” goals represents an inevitable choice for enterprises seeking high-quality development, with supply chain finance serving as an innovative financing tool that empowers businesses to embrace green innovation. This study employs data from Chinese A-share listed companies between 2010 and 2021, utilizing a multi-period difference-in-differences model to examine the impact of supply chain finance on corporate carbon performance from the dual perspectives of digital transformation and corporate value. The findings reveal: 1) Supply chain finance significantly enhances corporate carbon performance; 2) promoting digital transformation and elevating corporate value constitute the key mechanisms through which supply chain finance improves carbon performance; 3) heterogeneity analysis indicates that the carbon performance enhancement effect is more pronounced among large-scale enterprises, state-owned enterprises, and those with higher ESG ratings when adopting supply chain finance. These conclusions remain robust after testing for placebo effects, PSM-DID, and substitution variables. Machine learning validation confirms that incorporating SCF substantially improves predictive accuracy for corporate carbon performance. This study provides new micro-level evidence that developing SCF can enhance carbon reduction effects theoretically, while offering practical strategies for advancing and improving market-based green development.
Institutions play a crucial role in regulating and incentivizing corporate ESG disclosure practices. Existing studies have explored the mechanisms and pathways through which institutional elements affect the quality of corporate ESG disclosures. However, the understanding of these institutional elements remains somewhat narrow and linear, with limited consideration of the integrated effects of multiple institutional factors, as well as the potential differences in their significance and the possibility of nonlinear influences. Considering this, the present study, grounded in institutional theory, uses a sample of Chinese A-share listed companies from 2011 to 2021, and integrates various machine learning algorithms, including CatBoost, with SHAP values. The aim is to reveal the complex effects of diverse institutional factors, such as regulatory and normative elements, on the quality of corporate ESG disclosures. The results show that among the institutional elements, cultural-cognitive factors have the most significant impact on ESG disclosure quality. Previous disclosure quality emerges as the most important characteristic influencing corporate ESG disclosure quality. Within the various institutional elements, regional governance, public attention, and industry level are found to be the most influential in the regulatory, normative, and cultural-cognitive dimensions, respectively. Furthermore, most institutional characteristics exhibit nonlinear relationships with ESG disclosure quality. Further analysis demonstrates that the impact of institutional factors varies across regions, industries, and enterprise sizes. This study expands research on corporate ESG disclosure within the Chinese context, offering valuable theoretical insights and practical implications for government agencies in optimizing institutional frameworks and for companies aiming to enhance their ESG practices.
Based on 2007–2015 tax survey data, supply chain information of listed companies and city level data, this paper uses double machine learning and event study methods to analyze the impact of supply chain participation on the total factor productivity of SMEs. The empirical results show that after SMEs become the main customers of listed companies, the total factor productivity of SMEs will have a significant and sustained growth. By examining the performance of SMEs in terms of factor input and output, we find that supply chain participation increases labor input and output of enterprises, but the increase of factor input is smaller than the increase of output, so the total factor productivity increases. Furthermore, heterogeneity analysis reveals that the impact of supply chain participation is greater for capital-intensive enterprises, those in highly competitive industries, and high-tech industries. Finally, the mechanism analysis finds that the input effect of intermediate goods, the improvement of management ability and the alleviation of financing constraints are the important channels.
This paper aims to explore how supply chain members formulate optimal information acquisition and sharing strategies in scenarios where multiple quality information is unknown. To this end, this paper constructs a Stackelberg game model led by upstream enterprises, deeply analyzes the information acquisition decision-making behavior of upstream and downstream enterprises when facing unknown quality information, and discusses the potential impact of the probability and accuracy of high and low product quality signals on this decision-making process. The research results indicate that when the probability of high product quality is high, supply chain members tend to actively acquire information; however, when this probability is low, they tend not to acquire it. When the probability of high-quality signals is moderate, the equilibrium strategy for information acquisition by upstream and downstream enterprises varies with changes in the probability of high-quality signals. In addition, when the probability of high-quality signals is low, although upstream enterprises have an incentive to share the quality information they acquire, this can result in profit losses for downstream enterprises, who are therefore unwilling to accept this information. In such scenarios, the sharing of inaccurate information may lead to a decrease in overall supply chain profits, which contrasts with the conclusion from existing research that accurate information sharing enhances total supply chain profits. Nevertheless, when the probability of high-quality signals is high, through the implementation of revenue-sharing contracts, supply chain members can achieve mutual profit growth.
Customer resilience is a critical survival factor for companies in responding to market fluctuations and supply chain disruptions. Within corporate social responsibility (CSR) practices, product responsibility towards customers constitutes a core component, encompassing the safe design of products, provision of usage education, and ensuring product traceability. Grounded in social exchange theory, this study analyzes how corporate product responsibility practices enhance customer resilience. Through an empirical analysis of data from 200 Chinese manufacturing firms, the findings reveal that product responsibility practices effectively reduce customer opportunistic behavior, thereby enhancing customer resilience. Moreover, the study identifies the differential moderating roles of distributive and procedural justice asymmetries in the main relationship. Specifically, when there is an asymmetry in distributive justice, the impact of product responsibility practices on reducing customer opportunistic behavior is weaker, as customers are more inclined to maximize their own benefits, making it difficult to meet their economic needs. Conversely, when there is an asymmetry in procedural justice, the effect of product responsibility practices is stronger because the company’s social efforts can compensate for procedural injustices, satisfying customers’ need for a stable exchange process. These findings offer profound insights into how firms can build more resilient supply chains through their product responsibility practices, and they contribute to the literature on corporate social responsibility, supply chain resilience, and supply chain justice.
The coordinated development of enterprise basic research and enterprise performance is an integral part of strengthening the position of enterprise as the main body of scientific and technological innovation and achieving high-level scientific and technological self-reliance and self-improvement. Based on the panel data of Chinese industrial enterprises from 2009 to 2022, the study comprehensively utilizes the coupling coordination degree model, spatial econometric method, and structural equation model to systematically reveal the spatio-temporal evolution law and mechanism of the coupling coordination relationship between enterprise basic research and enterprise performance. The results show that 1) the coupling coordination degree of enterprise basic research and enterprise performance in China shows an increasing trend year by year, and is currently in the stage of high-quality coordination; 2) there is a significant spatial correlation in the coupling coordination degree between enterprise basic research and enterprise performance, and there is a negative coordination relationship between adjacent regions, with a spatial distribution pattern of clustering in dissimilar regions and dispersion in similar regions; 3) the regional differences in the coupling coordination degree between enterprise basic research and enterprise performance are showing a downward trend, mainly due to inter-regional differences, which are reflected in the eastern and northeastern regions, the central and northeastern regions, and the western and northeastern regions; 4) there is a significant positive coupling coordination effect between enterprise basic research and enterprise performance, and the main pathways include the basic research driven pathway, the technological innovation driven pathway, the production factor driven pathway, and the deep integration driven pathway. The research conclusions provide a theoretical basis for government departments to formulate innovation policies and for enterprises to optimize the allocation of research and development resources.
Influencers have become a crucial channel for brand promotion due to their personalized content and extensive influence. However, in pursuit of maximizing profits, some influencers selectively disclose information or overstate product benefits, compromising consumer rights and market integrity. This paper explores the impact of false advertising by influencers on business decision-making and product market performance in the context of influencer marketing. We use the finite period Bass diffusion model, combined with the advertising effect, to analyze the decision-making of false claims and its impact on product pricing and profits. The study reveals several insights. Firstly, once the sales cycle is established, the product’s sales rate becomes a constant. Secondly, while false claims can stimulate sales in the short term, in the long run, its negative repercussions outweigh initial gains. Particularly for brand manufacturer, the damage to reputation from false claims far exceeds short-term profit increases. In contrast, regular manufacturer are more inclined to adopt false advertising strategies to quickly capture market share. Lastly, businesses that engage in false claims tend to set higher product prices, yet their sales volume is not always higher. The contribution of this study is the novel application of the Bass model to the context of false advertising decision-making. From the perspective of merchants, this research systematically examines the decision-making mechanisms and influencing factors underlying false advertising practices. The findings highlight the potential adverse effects of false advertising strategies on long-term business profitability, offering a new theoretical framework for understanding corporate marketing decision-making behavior.
Based on the survey of the manufacturing practice of Seru by the globally leading monitor manufacturing enterprise in China, it is found that workers are affected by the learning effect during the assembly process, which may lead to the violation of constraints by the original assembly schedule. Therefore, considering the learning effect in the production process and minimizing the total quadratic completion time as the objective, this paper studies the scheduling problem of Yatai (single-person) Seru production. The problem is characterized by establishing a mixed integer programming model, and the objective function is linearized so that the model can be efficiently solved. The complexity of the problem is analyzed, and the optimal scheduling characteristics under the influence of the learning effect are proved. Then, an improved exact branch and bound algorithm and an improved tabu algorithm are designed according to the characteristics. The effectiveness of the algorithms is verified based on actual production and simulation data. In small scale experiments, the improved branch and bound algorithm proposed in this paper can obtain the optimal solution only by spending 0.4 times the CPU time of the Gurobi solver. In practical scale experiments, the improved tabu algorithm can solve examples with a scale of 15 workers and500 work orders within 400 seconds, and the time spent is about 1/3 of that of the classic tabu search algorithm. In addition, this paper also conducts sensitivity analysis experiments on the learning curve coefficient and analyzes the key factors of experimental parameters. The results show that the number of workers and the number of work orders are the main factors affecting the objective function. Therefore, increasing the number of Serus may be a more favorable management method than improving the workers’ learning effect.
In the context of China’s current structural tax reduction policy and common prosperity, it is of great significance to explore the impact of the tax reduction policy on the labor income share of enterprises. Using the tax investigation data from the Chinese state administration of tax (SAT), this paper explores the heterogeneous impact of China’s tax cut preferences for high-tech enterprises on their labor income shares based on a novel Bunching-Difference in Difference (Bunching-DiD) identification framework. It is found that tax cuts will prompt enterprises below the R&D intensity recognition standard to form clusters at the policy threshold, but without labor income share enhancement effect, while for the target of the policy—high-tech enterprises, they enhance R&D and innovation, increase the average labor wage, alleviate financing constraints, and reduce financing costs with the incentive of tax cuts, which has a significant labor income share enhancement effect. Secondly, this conclusion is robust after a series of robustness tests inside and outside the bunching intervals using suboptimal control intervals, replacing the sample of small firms, replacing the measure method, Bunching estimation calibration tests, respectively. Finally, the positive impact of tax incentives on the share of labor income is more significant in enterprises with lower industry concentration and state-owned enterprises, with higher fiscal pressure, lower level of marketization, stronger tax collection and management, and in the central and western regions. The research in this paper not only has certain value for understanding the relationship between enterprise tax reduction incentives and labor income share, but also provides decision-making references for improving the tax incentive policy for high-tech enterprise identification, promoting the construction of an innovative country and realizing the goal of common wealth.
“Challenges of risk assessment” has been a weak link in credit risk management of agricultural loans and a major constraint on the effective delivery of inclusive financial services to rural revitalization. The key to risk assessment lies in establishing effective identification criteria and adopting scientific methods for accurate evaluation. Taking the farmland management right mortgage as an example, this paper empirically examines the mitigating effects and implications of improvements in personal credit risk profiling and default identification methods on the “challenges of risk management” of agricultural loans, based on a comparative analysis of traditional and improved indicator systems and methods in personal credit risk identification. Results show that, in identifying personal credit risks, the new indicator system incorporating the “soft environment” of the agricultural loan market and farmers’ “soft information” demonstrates an information advantage over traditional indicator systems. Compared to traditional credit evaluation methods, the application of high-dimensional data-driven machine learning methods highlights a methodological advantage, with the methodological advantage outweighing the information advantage. Moreover, the new indicator system and method have the significant advantages in improving the accuracy of credit evaluation, which further leads to the reduction of loss measurement bias, the efficiency optimization of financial capital utilization, and the improvement of financial inclusion.
Accurately evaluating fund performance is a key issue in supporting the national strategy of cultivating world–class asset management institutions. Traditional asset pricing models, which rely on explicit factors, are insufficient in the era of big data where high–dimensional characteristics dominate fund performance. This paper employs the instrumented principal component analysis (IPCA) to extract latent pricing factors from 28fund–related characteristics. The resulting five–factor model achieves explanatory power of 81.85% and 99.82% in–sample, and 79.53% and 99.74% out–of–sample, for individual funds and fund portfolios, respectively. The model outperforms traditional factor models and mainstream machine learning algorithms in various dimensions. It demonstrates strong explanatory ability for both monthly and daily returns. The pricing power of the latent factors primarily stems from characteristics such as market capitalization, turnover rate, operating profitability, past performance of fund holdings, and historical fund returns, while also being influenced by macroeconomic variables including inflation, interest rates, macro leverage, market liquidity, and industrial production uncertainty. Based on these findings, the paper advocates for the development of a diversified, data–driven fund performance evaluation system to enhance retail investors’ asset allocation efficiency.
Data have become a crucial production factor for the digital economy and new productive forces, with the value increasingly recognized. Does this imply that data-related information disclosed by companies holds significant value? This study constructs a data element textual index from A-share firms’ annual reports and finds: 1) Higher data element content is linked to lower financing constraints, suggesting that more data information in reports helps ease financing issues; 2) this effect is stronger in state-owned firms, eastern regions, competitive industries, low-profit sectors, and high-tech fields; 3) enhanced data disclosure can reduce financing constraints by improving transparency, boosting innovation, and securing tax benefits. The study encourages firms to release data elements to promote digital transformation and productivity.
Measuring the contribution of institutional risk to systemic risk from the perspective of risk correlation is helpful to evaluate the effectiveness of decision-making in prudential supervision. Based on marginal expected shortfall (MES) and Bayesian network (BN), we propose a method to measure systemic risk represented by joint probability distribution of financial institution risks. An MES-BN risk dependence network is constructed to identify the institutional combinations that contribute most to systemic risks. Our main contribution is to introduce conditional dependent probability as a nonlinear feature of risk association, and identify the institution groups that contribute the most to systemic risk from the probability distribution of risk dependence, so as to provide a reference for the scope of supervision of systemic risk. During the financial crisis in 2008, the leveraged bull-bear period in2015 and the trade friction in 2019, the empirical results of 32 listed financial institutions in China show that: 1) Compared with the systemic risk without considering the risk dependence, the systemic risk based on the MES-BN risk dependence network can accurately measure the systemic risk from both the temporal similarity and the shape similarity; 2) the contribution MSR of the institutional combination based on the MES-BN model to systemic risk is
The interval Shapley value is a generalization of the Shapley value of the classic cooperative game when the payoffs are interval numbers. Most of the existing literature defines and characterizes the interval Shapley value based on the Hukuhara difference of interval numbers. This paper studies interval cooperative games with permission structures. First, we define the interval Shapley value based on the extension operations of interval numbers, and then use the interval Shapley values of restricted cooperative games to define the solution of such cooperative games. It is proved that this value is the only solution that satisfies additivity, central efficiency, inessential player, structural monotonicity, necessary player, marginal contributions, and zero solution. Finally, an example is given to illustrate the application of the solution to the problem of corporate profit distribution.
In real alliances, it is common for information to be incomplete. Meanwhile, the jealousy of the players will be more obvious due to the extreme uncertainty in the incomplete information, which affects the cooperation. This paper proposes a cooperative game with incomplete information and its solution considers players’ jealousy. Firstly, we define the double reference points based on the maximum and minimum income value that players may obtain in the alliance, and analyze the effect of jealousy on double reference points. Secondly, aim to maximize players’ distribution, an optimization mathematical model is constructed based on the sum of squares of the deviation between the distribution value and the jealous double reference points. Then the existence and properties of the optimal solution are proved. Finally, the proposed solution is applied to the profit distribution of inter-regional watershed joint development with incomplete information. Through comparative analysis, it is verified that this model is effective and feasible. This paper shows that jealousy will have different effects on players and negative effects on alliances.
The article considers cooperative games with transferable utility TU-games with graph structures and priority structures, it proposes the graph priority value which distributes each coalition’s graph Harsanyi dividend equally to the priority players in the coalition. In order to characterize this allocation rule, two pairs of axioms are proposed, involving those in the classical Shapley value and Myerson value, as well as new axioms for graph structures and priority structures. Finally, the article compares the relationship between the Owen value and the graph priority value, and analyses the differences between the Myerson value, the Shapley value, the priority value and the graph priority value by means of an arithmetic example.
When there are too many criteria involved in multi-criteria sorting problems, limited by the cognitive ability, decision makers (DMs) only focus on the performance of the alternatives on several criteria. Based on the disaggregation paradigm and regularization theory, we propose a novel multi-criteria sorting method that considers criteria selection. This method takes the sorting information of alternatives and capacity information of categories as input, and uses the complexity of the value function and its ability to restore the input information to guide the selection of the sorting model (including the criteria for evaluating alternatives, the value function, and the category thresholds). A 0–1 variable is introduced to indicate whether a criterion is selected or not, and a learning model is constructed to infer the values of parameters of the sorting model from the input information. In the obtained sorting model, the value function maintains the parsimonious form while ensuring the ability to restore the input information. If there is a value function constructed in accordance with a criteria subset (called the supporting criteria set) that can restore all input information, an iterative method is proposed to determine all supporting criteria sets that do not contain unnecessary criteria. This further enables us to analyze the possibility of a criterion being considered by DMs for evaluating alternatives. Finally, the effectiveness of the proposed method is illustrated by applying it to a problem of assessing debentures’ credit risk.
Recommendation systems are often affected by highly skewed long-tail item distributions in user-item interactions, where most long-tail items suffer from a lack of user feedback, significantly impacting recommendation performance. To improve the recommendation performance for long-tail items, long-tail item knowledge learned by recommendation models heavily rely on acquiring from head item knowledge on item side. However, most existing recommendation models for long-tail items tend to improve long-tail item knowledge learning at the expense of performance for head items. To do this, this paper proposes a decoupled knowledge transfer method for long-tail item recommendation. The method primarily focuses on learning from the long-tail item perspective, supplemented by learning from the head item perspective, achieving independent and complementary knowledge transfer through combining global and local graph structure learning methods. The main perspective focuses on knowledge transfer between head and long-tail items. Meanwhile, the supplementary perspective emphasizes knowledge reinforcement among head items, and guides the main perspective’s learning of head item knowledge. Additionally, the model gradually shifts its focus from the supplementary perspective to the main perspective as the training cycles progress by introducing a dynamic attention mechanism, achieving comprehensive learning of both head and long-tail item knowledge. Finally, experimental results on real-world datasets with long-tail distributions demonstrate that the recommendation model instantiated by the proposed method significantly improves the recommendation performance for long-tail items, while maintaining the performance for head items.
