Jin Xu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Jin Xu has a total of 30 co-authors. Here are the top co-authors by number of shared publications:
- Jinfeng Bu
School of Economics and Management, Southwest Jiaotong University, Chengdu, China
3 shared publications Tianrui Su
3 shared publicationsTianjing Yang
2 shared publicationsYing Jiang
Department of Information Systems and Operation Management, School of Economics and Management, Southwest Jiaotong University, Chengdu, China
2 shared publicationsYue Ding
2 shared publications- Andrianarivo Andriandafiarisoa Ralison Ny Avotra
Business School, Zhejiang Wanli University, Ningbo, China
1 shared publication
Latest Publications
- 20 August 2026
- Published by Springer Nature in Scientometrics
- Vol. 131 (10) , 7061-7099
Identifying multi-technology convergence patterns associated with emerging technologies: the case of new energy vehicles
- 20 August 2026
- Published by Springer Nature in Scientometrics
- Vol. 131 (10) , 7061-7099
Identifying multi-technology convergence patterns associated with emerging technologies: the case of new energy vehicles
- Open Access
- 12 May 2026
- Published by Elsevier in SSRN Electronic Journal
Community Detection in Passenger Hypernetworks and Prediction of Travel Mode Choice Based on Long-Distance joint travel
Railway and air travel have become the preferred modes of transport for long-distance travelers. Studying passengers’ travel characteristics and accurately predicting their mode preferences are crucial for improving passenger service standards. This study utilizes integrated transport big data to investigate the travel mode choices of long-distance travelers in depth. Firstly, the study constructs a passenger relationship hypernetwork that incorporates joint travel relationships; secondly, a two-stage detection algorithm is proposed to identify closely associated groups of travelers; Subsequently, a passenger activity-influence model was established to predict travel mode choices. Finally, the validity of this model was verified through an empirical study utilizing railway and aviation data from south-western China. The results indicate that the proposed community detection algorithm possesses robust cluster identification capabilities and resistance to interference. By combining passenger characteristics with network topological features, the model significantly improved the accuracy of travel mode preference predictions, correctly forecasting the travel mode preferences of over 92.09% of passengers.
- 12 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessCommunity Detection in Passenger Hypernetworks and Prediction of Travel Mode Choice Based on Long-Distance joint travel
Railway and air travel have become the preferred modes of transport for long-distance travelers. Studying passengers’ travel characteristics and accurately predicting their mode preferences are crucial for improving passenger service standards. This study utilizes integrated transport big data to investigate the travel mode choices of long-distance travelers in depth. Firstly, the study constructs a passenger relationship hypernetwork that incorporates joint travel relationships; secondly, a two-stage detection algorithm is proposed to identify closely associated groups of travelers; Subsequently, a passenger activity-influence model was established to predict travel mode choices. Finally, the validity of this model was verified through an empirical study utilizing railway and aviation data from south-western China. The results indicate that the proposed community detection algorithm possesses robust cluster identification capabilities and resistance to interference. By combining passenger characteristics with network topological features, the model significantly improved the accuracy of travel mode preference predictions, correctly forecasting the travel mode preferences of over 92.09% of passengers.
- Open Access
- 12 May 2026
- Published by Elsevier in SSRN Electronic Journal
Community Detection in Passenger Hypernetworks and Prediction of Travel Mode Choice Based on Long-Distance joint travel
Railway and air travel have become the preferred modes of transport for long-distance travelers. Studying passengers’ travel characteristics and accurately predicting their mode preferences are crucial for improving passenger service standards. This study utilizes integrated transport big data to investigate the travel mode choices of long-distance travelers in depth. Firstly, the study constructs a passenger relationship hypernetwork that incorporates joint travel relationships; secondly, a two-stage detection algorithm is proposed to identify closely associated groups of travelers; Subsequently, a passenger activity-influence model was established to predict travel mode choices. Finally, the validity of this model was verified through an empirical study utilizing railway and aviation data from south-western China. The results indicate that the proposed community detection algorithm possesses robust cluster identification capabilities and resistance to interference. By combining passenger characteristics with network topological features, the model significantly improved the accuracy of travel mode preference predictions, correctly forecasting the travel mode preferences of over 92.09% of passengers.
- 12 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessCommunity Detection in Passenger Hypernetworks and Prediction of Travel Mode Choice Based on Long-Distance joint travel
Railway and air travel have become the preferred modes of transport for long-distance travelers. Studying passengers’ travel characteristics and accurately predicting their mode preferences are crucial for improving passenger service standards. This study utilizes integrated transport big data to investigate the travel mode choices of long-distance travelers in depth. Firstly, the study constructs a passenger relationship hypernetwork that incorporates joint travel relationships; secondly, a two-stage detection algorithm is proposed to identify closely associated groups of travelers; Subsequently, a passenger activity-influence model was established to predict travel mode choices. Finally, the validity of this model was verified through an empirical study utilizing railway and aviation data from south-western China. The results indicate that the proposed community detection algorithm possesses robust cluster identification capabilities and resistance to interference. By combining passenger characteristics with network topological features, the model significantly improved the accuracy of travel mode preference predictions, correctly forecasting the travel mode preferences of over 92.09% of passengers.
- Open Access
- 9 May 2026
- Published by Elsevier in SSRN Electronic Journal
How Investor Position Influences Innovation: A Complex Network Perspective
This study empirically examines the impact pathways of investors' network positions on corporate innovation from a micro-level perspective. Findings reveal the following: First, an investor's centrality position within the network significantly and positively influences the patent output of its portfolio firms. Second, the impact of network position on innovation collaboration strategies is more complex: an investor's proactive outward connections actually inhibit the breadth and depth of corporate innovation. More critically, the reputation and activity level of an investor's direct neighbors also exert a significant negative influence on the breadth and depth of corporate collaboration, revealing that high-quality partners may trigger relationship lock-in effects. Third, the breadth and depth of innovation collaboration serve as key mediating channels through which network position influences innovation output, with the mediating effect of collaboration depth being particularly crucial. Fourth, these mechanisms are moderated by internal and external contextual factors: the heterogeneity of joint investment teams significantly amplifies the innovation returns from network position advantages, while a favorable market environment weakens the effectiveness of proactive connection strategies. Further causal analysis indicates that macro-level innovation policies significantly enhance innovation output, but their impact primarily relies on direct empowerment rather than restructuring collaborative networks.
- 9 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessHow Investor Position Influences Innovation: A Complex Network Perspective
This study empirically examines the impact pathways of investors' network positions on corporate innovation from a micro-level perspective. Findings reveal the following: First, an investor's centrality position within the network significantly and positively influences the patent output of its portfolio firms. Second, the impact of network position on innovation collaboration strategies is more complex: an investor's proactive outward connections actually inhibit the breadth and depth of corporate innovation. More critically, the reputation and activity level of an investor's direct neighbors also exert a significant negative influence on the breadth and depth of corporate collaboration, revealing that high-quality partners may trigger relationship lock-in effects. Third, the breadth and depth of innovation collaboration serve as key mediating channels through which network position influences innovation output, with the mediating effect of collaboration depth being particularly crucial. Fourth, these mechanisms are moderated by internal and external contextual factors: the heterogeneity of joint investment teams significantly amplifies the innovation returns from network position advantages, while a favorable market environment weakens the effectiveness of proactive connection strategies. Further causal analysis indicates that macro-level innovation policies significantly enhance innovation output, but their impact primarily relies on direct empowerment rather than restructuring collaborative networks.
- 11 September 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 298
Companion learning Networks: A deep reinforcement learning algorithm with partner networks
- 11 September 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 298
Companion learning Networks: A deep reinforcement learning algorithm with partner networks
- 22 August 2024
- Published by SAGE Publications in Transportation Research Record: Journal of the Transportation Research Board
- Vol. 2679 (2) , 1344-1358
Analysis of Travel Mode Choice Behavior between High-Speed Rail and Air Transport Utilizing Large-Scale Ticketing Data
As essential infrastructure, high-speed rail (HSR) and air transport (AT) play crucial roles in socioeconomic development. With their continuous expansion in China, the overlap of HSR and AT networks has increased, providing travelers with more choices for intercity travel. Because fierce competition in the medium-to-long-distance segment affects the market share and transport capacity dispatching, the travel choice between HSR and AT has been of intense interest. This study utilized a unique fusion dataset collected from two separate organizations to conduct an empirical analysis of the travel mode choice behaviors of individuals when choosing between HSR and AT. A multinomial logit (MNL) model was adopted to examine the influences of key factors on passenger choice preferences. The results showed that the fitting effect of the MNL model was satisfactory, and the parameters were strongly interpretable. The McFadden Pseudo R 2 with a city-pair fixed effect in the MNL model increased by 17.3% compared with that without the city-pair fixed effect. All the related explanatory variables, including the trip distance by high-speed train, demography, ticket purchasing, and travel behavior characteristics, had significant positive effects on the passengers’ choice of AT, with trip distance having the largest effect. According to the parameter estimation, 1,160 km was the division for individual choice between HSR and AT. This study also compared the prediction accuracies of the MNL model and eight classical machine-learning models and found that random forest had the best performance. This study provides a new framework for analyzing travel choice modeling when choosing between HSR and AT.
- 22 August 2024
- Published by SAGE Publications in Transportation Research Record: Journal of the Transportation Research Board
- Vol. 2679 (2) , 1344-1358
Analysis of Travel Mode Choice Behavior between High-Speed Rail and Air Transport Utilizing Large-Scale Ticketing Data
As essential infrastructure, high-speed rail (HSR) and air transport (AT) play crucial roles in socioeconomic development. With their continuous expansion in China, the overlap of HSR and AT networks has increased, providing travelers with more choices for intercity travel. Because fierce competition in the medium-to-long-distance segment affects the market share and transport capacity dispatching, the travel choice between HSR and AT has been of intense interest. This study utilized a unique fusion dataset collected from two separate organizations to conduct an empirical analysis of the travel mode choice behaviors of individuals when choosing between HSR and AT. A multinomial logit (MNL) model was adopted to examine the influences of key factors on passenger choice preferences. The results showed that the fitting effect of the MNL model was satisfactory, and the parameters were strongly interpretable. The McFadden Pseudo R 2 with a city-pair fixed effect in the MNL model increased by 17.3% compared with that without the city-pair fixed effect. All the related explanatory variables, including the trip distance by high-speed train, demography, ticket purchasing, and travel behavior characteristics, had significant positive effects on the passengers’ choice of AT, with trip distance having the largest effect. According to the parameter estimation, 1,160 km was the division for individual choice between HSR and AT. This study also compared the prediction accuracies of the MNL model and eight classical machine-learning models and found that random forest had the best performance. This study provides a new framework for analyzing travel choice modeling when choosing between HSR and AT.
- 10 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Engineering Management
- Vol. 71 (00189391) , 11361-11374
A Knowledge Transfer Framework Based on Deep-Reinforcement Learning for Multistage Construction Projects
Construction projects can produce excessive construction data using intelligent equipment. Deep learning algorithms can harness these data, discovering knowledge that can effectively enhance project execution and performance. However, these algorithms' efficacy hinges on the availability of data, often presenting a challenge at project initiation when data are scarce or absent. Although this issue can be alleviated through knowledge transfer algorithms, there is still a lack of a method for selecting knowledge source domains that can adapt to data distributions. Further, the long durations and dynamic environments of construction projects require timely updates to a project's knowledge base, which is often ignored by current knowledge-management practices. This study introduces a multi-stage, online knowledge transfer framework tailored to three distinct stages of construction projects: the initiation, data-emergence, and data-rich stages. Firstly, the framework transforms sourcQe-project data into target-project data, alleviating the initial data deficit. Subsequently, it utilizes a combination of multiple similarity metrics and a deep-reinforcement learning model to adaptively select source domains with minimal distribution disparities, improving the effect of knowledge transfer. Finally, it integrates concept-drift algorithms and constructs knowledge-source discrimination rule to automate the selection of knowledge sources and schedule updates, enabling dynamic knowledge transfer in construction projects.
- 10 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Engineering Management
- Vol. 71 (00189391) , 11361-11374
A Knowledge Transfer Framework Based on Deep-Reinforcement Learning for Multistage Construction Projects
Construction projects can produce excessive construction data using intelligent equipment. Deep learning algorithms can harness these data, discovering knowledge that can effectively enhance project execution and performance. However, these algorithms' efficacy hinges on the availability of data, often presenting a challenge at project initiation when data are scarce or absent. Although this issue can be alleviated through knowledge transfer algorithms, there is still a lack of a method for selecting knowledge source domains that can adapt to data distributions. Further, the long durations and dynamic environments of construction projects require timely updates to a project's knowledge base, which is often ignored by current knowledge-management practices. This study introduces a multi-stage, online knowledge transfer framework tailored to three distinct stages of construction projects: the initiation, data-emergence, and data-rich stages. Firstly, the framework transforms sourcQe-project data into target-project data, alleviating the initial data deficit. Subsequently, it utilizes a combination of multiple similarity metrics and a deep-reinforcement learning model to adaptively select source domains with minimal distribution disparities, improving the effect of knowledge transfer. Finally, it integrates concept-drift algorithms and constructs knowledge-source discrimination rule to automate the selection of knowledge sources and schedule updates, enabling dynamic knowledge transfer in construction projects.
- 27 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Engineering Management
- Vol. 71 (00189391) , 9938-9952
An R&D Partner Recommendation Framework Based on a Knowledge Context Hypernetwork for Engineering Technological Innovation
Engineering construction enterprises often need to partner with R&D institutions and jointly carry out technological innovation activities. Identifying the right R&D partner(s) for the job is key. An accurate selection of R&D partners requires examining the knowledge gained from their past innovations and the corresponding contexts. This article introduces a novel R&D partner recommendation framework (PROKCH) based on a knowledge context hypernetwork to analyze the knowledge contexts of engineering technology innovation for recommending R&D partners. First, in this article, eight dimensions of knowledge contexts are defined, and an automated method to extract them from related documents is proposed. Next, a four-layer hypernetwork model is constructed to represent the extracted knowledge contexts. Finally, we develop a novel algorithm based on node similarity in the hypernetwork for recommending R&D partners. Experiments on a real dataset of recent railway tunneling projects in China demonstrate that the proposed PROKCH framework can improve the performance of R&D institution recommendations.
- 27 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Engineering Management
- Vol. 71 (00189391) , 9938-9952
An R&D Partner Recommendation Framework Based on a Knowledge Context Hypernetwork for Engineering Technological Innovation
Engineering construction enterprises often need to partner with R&D institutions and jointly carry out technological innovation activities. Identifying the right R&D partner(s) for the job is key. An accurate selection of R&D partners requires examining the knowledge gained from their past innovations and the corresponding contexts. This article introduces a novel R&D partner recommendation framework (PROKCH) based on a knowledge context hypernetwork to analyze the knowledge contexts of engineering technology innovation for recommending R&D partners. First, in this article, eight dimensions of knowledge contexts are defined, and an automated method to extract them from related documents is proposed. Next, a four-layer hypernetwork model is constructed to represent the extracted knowledge contexts. Finally, we develop a novel algorithm based on node similarity in the hypernetwork for recommending R&D partners. Experiments on a real dataset of recent railway tunneling projects in China demonstrate that the proposed PROKCH framework can improve the performance of R&D institution recommendations.
- 23 May 2023
- Published by Emerald Publishing in Business Process Management Journal
- Vol. 29 (4) , 1249-1276
Does leadership and management support facilitate green knowledge acquisition and green innovation: a moderated mediation approach
Purpose: Given the critical role of green innovation (GI) in the manufacturing sector, this study builds a moderated mediation model to evaluate the influence of leadership and management support on GI, the mediating function of green knowledge acquisition, and the moderating role of green absorptive ability.Design/methodology/approach: The study employed a quantitative research approach with hierarchical regression analysis to assess the proposed relationships among the constructs on a sample of 371 executives from 117 large-sized manufacturing firms in Pakistan.Findings: The research findings demonstrate that leadership and management support significantly affects both radical and incremental GI, with incremental green innovation being more positively affected than radical green innovation. Green knowledge acquisition partially mediates between leadership and management support, radical and incremental green innovation. Green knowledge acquisition moderates the association between leadership and management support and green knowledge acquisition and the link between leadership and management support and incremental GI. The findings also demonstrate that green knowledge acquisition's mediating effect on leadership and management support, and GI is more pronounced when green absorptive capacity is high.Research limitations/implications: This research is based on cross-sectional data gathered from manufacturing companies. Future studies should consider this differentiation between the enterprises since there are various sectors within the general manufacturing sector whose environmental effect is more or less polluting. This research focused exclusively on two aspects of GI (radical and incremental GI). It is feasible that additional GI constituents (i.e., product, process, and management GI) can significantly boost businesses' competitive advantage. This study recommends additional study into the potential moderating impacts of technological and market turbulence to better understand the relationship between these concepts since it is evident that internal and external factors influence GI.Practical implications: The study provides useful insights and an innovative way for manufacturing firms and authorities to prevent environmental deterioration and achieve sustainable green innovation through leadership and management support and green intangible resources.Originality/value: This research concentrating on green environmental concerns and using RBV theory attempts to fill research gaps and sheds light on how leadership and management support promote both radical and incremental green innovation via the mediating and moderating roles of green knowledge acquisition and green absorptive capacity.
- 23 May 2023
- Published by Emerald Publishing in Business Process Management Journal
- Vol. 29 (4) , 1249-1276
Does leadership and management support facilitate green knowledge acquisition and green innovation: a moderated mediation approach
Purpose: Given the critical role of green innovation (GI) in the manufacturing sector, this study builds a moderated mediation model to evaluate the influence of leadership and management support on GI, the mediating function of green knowledge acquisition, and the moderating role of green absorptive ability.Design/methodology/approach: The study employed a quantitative research approach with hierarchical regression analysis to assess the proposed relationships among the constructs on a sample of 371 executives from 117 large-sized manufacturing firms in Pakistan.Findings: The research findings demonstrate that leadership and management support significantly affects both radical and incremental GI, with incremental green innovation being more positively affected than radical green innovation. Green knowledge acquisition partially mediates between leadership and management support, radical and incremental green innovation. Green knowledge acquisition moderates the association between leadership and management support and green knowledge acquisition and the link between leadership and management support and incremental GI. The findings also demonstrate that green knowledge acquisition's mediating effect on leadership and management support, and GI is more pronounced when green absorptive capacity is high.Research limitations/implications: This research is based on cross-sectional data gathered from manufacturing companies. Future studies should consider this differentiation between the enterprises since there are various sectors within the general manufacturing sector whose environmental effect is more or less polluting. This research focused exclusively on two aspects of GI (radical and incremental GI). It is feasible that additional GI constituents (i.e., product, process, and management GI) can significantly boost businesses' competitive advantage. This study recommends additional study into the potential moderating impacts of technological and market turbulence to better understand the relationship between these concepts since it is evident that internal and external factors influence GI.Practical implications: The study provides useful insights and an innovative way for manufacturing firms and authorities to prevent environmental deterioration and achieve sustainable green innovation through leadership and management support and green intangible resources.Originality/value: This research concentrating on green environmental concerns and using RBV theory attempts to fill research gaps and sheds light on how leadership and management support promote both radical and incremental green innovation via the mediating and moderating roles of green knowledge acquisition and green absorptive capacity.