Minghu Tang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Minghu Tang has a total of 24 co-authors. Here are the top co-authors by number of shared publications:
- Junjie Wang
School of Intelligent Science and Engineering, Qinghai Minzu University, Xining 810007, China
3 shared publications - Peng Luo
College of Computer Science and Technology, Qinghai Normal University, Xining 810016, China
3 shared publications - Jiayi Zhang
School of Intelligent Science and Engineering, Qinghai Minzu University, Xining 810007, China
2 shared publications - Wei Yu
College of International Business, Zhejiang Yuexiu University, Shaoxing, China
2 shared publications Wenjun Wang
2 shared publications- Xiaoming Li
School of International Business, Zhejiang Yuexiu University, Shaoxing, China
2 shared publications
Latest Publications
- Open Access
- 23 October 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (21) , 10519
Enhanced Mask R-CNN with PANet and PointRend for Detection and Instance Segmentation of Shadow Puppet Characters
The segmentation of Hehuang shadow puppet images holds significant value for artistic creation and cultural preservation. However, it still faces challenges such as the subjective nature of manual segmentation and inaccurate boundary delineation in complex images. To address these challenges, an improved Mask R-CNN model is proposed for precise segmentation of shadow puppet image boundaries. Specifically, inspired by PANet, only its bottom-up path augmentation is adopted after the FPN outputs to shorten the feature fusion path between low-level contour features and high-level semantics. Additionally, Concat replaces residual connections to preserve feature diversity. Finally, the PointRend module is integrated into the mask branch to prevent overly smooth transitions along segmentation mask edges and enhance pixel-level refinement of image boundary segmentation. In this study, a novel COCO-format dataset with 2700 images and seven instance categories was constructed for comprehensive segmentation evaluation. Compared with the baseline, the proposed model achieves relative improvements of 10.1% in bbox_AP for detection and 10.6% in segm_AP for instance segmentation, while segm_AP50 rises from 92.1% to 95.0%. Furthermore, we developed an interactive segmentation system that refines complex boundaries, enhancing the extraction capability of shadow puppet elements and facilitating its application in digital preservation scenarios of shadow puppet heritage.
- 23 October 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (21) , 10519
Open AccessEnhanced Mask R-CNN with PANet and PointRend for Detection and Instance Segmentation of Shadow Puppet Characters
The segmentation of Hehuang shadow puppet images holds significant value for artistic creation and cultural preservation. However, it still faces challenges such as the subjective nature of manual segmentation and inaccurate boundary delineation in complex images. To address these challenges, an improved Mask R-CNN model is proposed for precise segmentation of shadow puppet image boundaries. Specifically, inspired by PANet, only its bottom-up path augmentation is adopted after the FPN outputs to shorten the feature fusion path between low-level contour features and high-level semantics. Additionally, Concat replaces residual connections to preserve feature diversity. Finally, the PointRend module is integrated into the mask branch to prevent overly smooth transitions along segmentation mask edges and enhance pixel-level refinement of image boundary segmentation. In this study, a novel COCO-format dataset with 2700 images and seven instance categories was constructed for comprehensive segmentation evaluation. Compared with the baseline, the proposed model achieves relative improvements of 10.1% in bbox_AP for detection and 10.6% in segm_AP for instance segmentation, while segm_AP50 rises from 92.1% to 95.0%. Furthermore, we developed an interactive segmentation system that refines complex boundaries, enhancing the extraction capability of shadow puppet elements and facilitating its application in digital preservation scenarios of shadow puppet heritage.
- Open Access
- 20 August 2026
- Published by Elsevier in Results in Engineering
- Vol. 32
NCTAF-GCN: Node-Contrastive Topology-Adaptive Fairing Graph Convolutional Network for anomaly detection
- 20 August 2026
- Published by Elsevier in Results in Engineering
- Vol. 32
Open AccessNCTAF-GCN: Node-Contrastive Topology-Adaptive Fairing Graph Convolutional Network for anomaly detection
- Open Access
- 9 April 2026
- Published by MDPI AG in Information
- Vol. 17 (4) , 360
Multi-Strategy Improvement and Comparative Research on Data-Driven Social Network Construction in Edge-Deficient Scenarios for Social Bot Account Detection
Accurate social bot detection relies on simulated data to alleviate the scarcity of labeled real-world datasets. Synthetic graph data serves as the core training resource for detection models within simulated data; nevertheless, edge deficiency in real social networks (induced by privacy constraints and data collection limitations) gives rise to “pseudo-isolated nodes” and distorts the quality of synthetic graph data. Furthermore, mainstream data-driven synthetic graph generation methods lack systematic and credible comparative analyses. To tackle these problems, this study optimizes two representative synthetic graph generation approaches (the Chung-Lu model and the Random Classifier-based Multi-Hop (RCMH) sampling + diffusion model) and puts forward an edge completion strategy grounded in sociological theories. Multiple groups of comparative experiments are conducted to assess the performance of the improved methods and the edge completion strategy. Experimental results demonstrate that the “interest + social association” edge completion strategy achieves an F1-score (F1) of 0.7051, and the improved sampling + diffusion model integrated with edge completion reaches an F1-score of 0.7071, which performs better than traditional and unmodified methods to a certain extent. This work preliminarily enhances the reliability of synthetic graph generation methods and provides relatively high-quality synthetic social graph data for social bot detection. It should be noted that the proposed methods are validated solely on Twitter-derived datasets, and their effectiveness remains to be verified in cross-platform adaptation and dynamic social network scenarios.
- 9 April 2026
- Published by MDPI AG in Information
- Vol. 17 (4) , 360
Open AccessMulti-Strategy Improvement and Comparative Research on Data-Driven Social Network Construction in Edge-Deficient Scenarios for Social Bot Account Detection
Accurate social bot detection relies on simulated data to alleviate the scarcity of labeled real-world datasets. Synthetic graph data serves as the core training resource for detection models within simulated data; nevertheless, edge deficiency in real social networks (induced by privacy constraints and data collection limitations) gives rise to “pseudo-isolated nodes” and distorts the quality of synthetic graph data. Furthermore, mainstream data-driven synthetic graph generation methods lack systematic and credible comparative analyses. To tackle these problems, this study optimizes two representative synthetic graph generation approaches (the Chung-Lu model and the Random Classifier-based Multi-Hop (RCMH) sampling + diffusion model) and puts forward an edge completion strategy grounded in sociological theories. Multiple groups of comparative experiments are conducted to assess the performance of the improved methods and the edge completion strategy. Experimental results demonstrate that the “interest + social association” edge completion strategy achieves an F1-score (F1) of 0.7051, and the improved sampling + diffusion model integrated with edge completion reaches an F1-score of 0.7071, which performs better than traditional and unmodified methods to a certain extent. This work preliminarily enhances the reliability of synthetic graph generation methods and provides relatively high-quality synthetic social graph data for social bot detection. It should be noted that the proposed methods are validated solely on Twitter-derived datasets, and their effectiveness remains to be verified in cross-platform adaptation and dynamic social network scenarios.
- Open Access
- 25 March 2026
- Published by MDPI AG in Information
- Vol. 17 (4) , 320
Detection of LLM-Generated Text vs. Human Text via DeepSeek-R1 Multi-Feature Fusion
The development of generative artificial intelligence technology has brought convenience to various industries, but also caused some confusion. Especially today, when the content generated by large language models is extremely similar to real text, it has created challenges in many fields (such as in the discrimination of graduation theses in schools) in quickly identifying whether a text is from human sources or generated by large language models. Based on the DeepSeek-R1 language model, this paper combines natural language features and uses a judgment mechanism to detect text generated by large language models. Experimental results show that its accuracy is improved compared with conventional methods in the Reuters, WP and HC3 datasets.
- 25 March 2026
- Published by MDPI AG in Information
- Vol. 17 (4) , 320
Open AccessDetection of LLM-Generated Text vs. Human Text via DeepSeek-R1 Multi-Feature Fusion
The development of generative artificial intelligence technology has brought convenience to various industries, but also caused some confusion. Especially today, when the content generated by large language models is extremely similar to real text, it has created challenges in many fields (such as in the discrimination of graduation theses in schools) in quickly identifying whether a text is from human sources or generated by large language models. Based on the DeepSeek-R1 language model, this paper combines natural language features and uses a judgment mechanism to detect text generated by large language models. Experimental results show that its accuracy is improved compared with conventional methods in the Reuters, WP and HC3 datasets.
- Open Access
- 24 March 2026
- Published by MDPI AG in Electronics
- Vol. 15 (7) , 1347
Exploration of Thangka Identification and Traceability Mechanism Empowered by Blockchain
Authenticity verification for thangka artworks remains challenging in the market, as traditional physical authentication methods offer limited reliability, while modern spectroscopic and chemical testing technologies are costly and unsuitable for large-scale application. Although deep learning methods can achieve efficient authentication through image features, they rely on centralized databases to store feature information, making them susceptible to tampering risks and undermining the credibility of authentication results. To address these issues, this study proposes a digital authentication method for thangka paintings that integrates blockchain technology. This approach stores image features in the InterPlanetary File System (IPFS) and records their hash values on the blockchain, ensuring the immutability and verifiable evidence of feature data. Simultaneously, it employs convolutional neural networks for feature extraction and similarity analysis of thangka images, constructing an integrated platform system encompassing storage, authentication, and traceability. This enhances the reliability and automation of authentication outcomes. The platform further supports full-process traceability of thangka storage and authentication operations, providing a viable pathway for establishing a scientific and reliable digital authentication system for thangkas. Experimental evaluation on a dataset of 2847 thangka images demonstrates 99.2% authentication accuracy, with a precision of 98.7% and an F1-score of 99.1%, while end-to-end authentication latency averages 1247 ms, validating the system’s effectiveness for practical museum and market deployment scenarios.
- 24 March 2026
- Published by MDPI AG in Electronics
- Vol. 15 (7) , 1347
Open AccessExploration of Thangka Identification and Traceability Mechanism Empowered by Blockchain
Authenticity verification for thangka artworks remains challenging in the market, as traditional physical authentication methods offer limited reliability, while modern spectroscopic and chemical testing technologies are costly and unsuitable for large-scale application. Although deep learning methods can achieve efficient authentication through image features, they rely on centralized databases to store feature information, making them susceptible to tampering risks and undermining the credibility of authentication results. To address these issues, this study proposes a digital authentication method for thangka paintings that integrates blockchain technology. This approach stores image features in the InterPlanetary File System (IPFS) and records their hash values on the blockchain, ensuring the immutability and verifiable evidence of feature data. Simultaneously, it employs convolutional neural networks for feature extraction and similarity analysis of thangka images, constructing an integrated platform system encompassing storage, authentication, and traceability. This enhances the reliability and automation of authentication outcomes. The platform further supports full-process traceability of thangka storage and authentication operations, providing a viable pathway for establishing a scientific and reliable digital authentication system for thangkas. Experimental evaluation on a dataset of 2847 thangka images demonstrates 99.2% authentication accuracy, with a precision of 98.7% and an F1-score of 99.1%, while end-to-end authentication latency averages 1247 ms, validating the system’s effectiveness for practical museum and market deployment scenarios.
- Open Access
- 28 December 2025
- Published by MDPI AG in Mathematics
- Vol. 14 (1) , 113
Effective Target Privacy Protection Against Dynamic-Link-Prediction-Based Attacks via Adversarial Learning
Graph data mining has emerged as a prominent area of research in both academic and industrial domains. Dynamic link prediction, a critical subfield within graph data mining, offers a more realistic representation of real-world networks compared to static link prediction, making dynamic link prediction attacks particularly threatening to privacy. While privacy protection in dynamic networks can be achieved by removing certain sensitive links, attackers can still infer hidden sensitive connections from observable network data. Moreover, existing studies seldom address target-level defense against dynamic link prediction attacks. To address these challenges, this paper proposes a Target-Level Privacy protection method against Dynamic Link Prediction attacks (TP-DLP). The method leverages temporal information in dynamic networks to implement targeted protection based on link gradient information, operating within a perturbation range that preserves the inherent characteristics of dynamic networks. Using dynamic social networks as a case study, the approach distinguishes the privacy levels of dynamic links to achieve target-level privacy protection. Extensive experimental results demonstrate that TP-DLP significantly enhances privacy protection while preserving network utility, making it well-suited for targeted defense against dynamic network link prediction. It can be concluded that our method achieves a balanced trade-off between privacy protection effectiveness and network structural fidelity.
- 28 December 2025
- Published by MDPI AG in Mathematics
- Vol. 14 (1) , 113
Open AccessEffective Target Privacy Protection Against Dynamic-Link-Prediction-Based Attacks via Adversarial Learning
Graph data mining has emerged as a prominent area of research in both academic and industrial domains. Dynamic link prediction, a critical subfield within graph data mining, offers a more realistic representation of real-world networks compared to static link prediction, making dynamic link prediction attacks particularly threatening to privacy. While privacy protection in dynamic networks can be achieved by removing certain sensitive links, attackers can still infer hidden sensitive connections from observable network data. Moreover, existing studies seldom address target-level defense against dynamic link prediction attacks. To address these challenges, this paper proposes a Target-Level Privacy protection method against Dynamic Link Prediction attacks (TP-DLP). The method leverages temporal information in dynamic networks to implement targeted protection based on link gradient information, operating within a perturbation range that preserves the inherent characteristics of dynamic networks. Using dynamic social networks as a case study, the approach distinguishes the privacy levels of dynamic links to achieve target-level privacy protection. Extensive experimental results demonstrate that TP-DLP significantly enhances privacy protection while preserving network utility, making it well-suited for targeted defense against dynamic network link prediction. It can be concluded that our method achieves a balanced trade-off between privacy protection effectiveness and network structural fidelity.
- Open Access
- 13 September 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (18) , 10028
Research on BBHL Model Based on Hybrid Loss Optimization for Fake News Detection
With the rapid development of social media, the spread of fake news has become a significant issue affecting social stability. To address the problems of incomplete feature extraction and simplistic loss function design in traditional fake news detection, this paper proposes a BBHL model based on hybrid loss optimization. The model achieves deep extraction of text features by integrating BERT, Bi-LSTM, and attention mechanisms, and innovatively fuses binary cross-entropy (BCE) loss with contrastive loss to enhance feature discriminability and the model’s generalization ability. Experiments on the Weibo, Twitter, and Pheme datasets demonstrate that the BBHL model significantly outperforms baseline models such as EANN and MCNN in metrics including accuracy and F1-score. Ablation experiments verify the effectiveness of contrastive loss, providing a robust and generalizable solution for fake news detection.
- 13 September 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (18) , 10028
Open AccessResearch on BBHL Model Based on Hybrid Loss Optimization for Fake News Detection
With the rapid development of social media, the spread of fake news has become a significant issue affecting social stability. To address the problems of incomplete feature extraction and simplistic loss function design in traditional fake news detection, this paper proposes a BBHL model based on hybrid loss optimization. The model achieves deep extraction of text features by integrating BERT, Bi-LSTM, and attention mechanisms, and innovatively fuses binary cross-entropy (BCE) loss with contrastive loss to enhance feature discriminability and the model’s generalization ability. Experiments on the Weibo, Twitter, and Pheme datasets demonstrate that the BBHL model significantly outperforms baseline models such as EANN and MCNN in metrics including accuracy and F1-score. Ablation experiments verify the effectiveness of contrastive loss, providing a robust and generalizable solution for fake news detection.
- Open Access
- 28 February 2025
- Published by MDPI AG in Mathematics
- Vol. 13 (5) , 807
A Review of Link Prediction Algorithms in Dynamic Networks
Dynamic network link prediction refers to the prediction of possible future links or the identification of missing links on the basis of historical information of dynamic networks. Link prediction aids people in exploring and analyzing complex change patterns in the real world and it could be applied in personalized recommendation systems, intelligence analysis, anomaly detection, and other fields. This paper aims to provide a comprehensive review of dynamic network link prediction. Firstly, dynamic networks are categorized into dynamic univariate networks and dynamic multivariate networks according to the changes in their sets. Furthermore, dynamic network link prediction algorithms are classified into regular sampling and irregular sampling by the method of network sampling. After summarizing and comparing the common datasets and evaluation indicators for dynamic network link prediction, we briefly review classic related algorithms in recent years, and classify them according to the network changes, sampling methods, underlying principles of algorithms, and other classification methods. Meanwhile, the basic ideas, advantages, and disadvantages of these algorithms are discussed in detail. The application fields and challenges in this area are also summarized. In the final summary of the paper, the future research directions such as link prediction in dynamic heterogeneous weighted networks and the security issues brought about by link prediction are discussed.
- 28 February 2025
- Published by MDPI AG in Mathematics
- Vol. 13 (5) , 807
Open AccessA Review of Link Prediction Algorithms in Dynamic Networks
Dynamic network link prediction refers to the prediction of possible future links or the identification of missing links on the basis of historical information of dynamic networks. Link prediction aids people in exploring and analyzing complex change patterns in the real world and it could be applied in personalized recommendation systems, intelligence analysis, anomaly detection, and other fields. This paper aims to provide a comprehensive review of dynamic network link prediction. Firstly, dynamic networks are categorized into dynamic univariate networks and dynamic multivariate networks according to the changes in their sets. Furthermore, dynamic network link prediction algorithms are classified into regular sampling and irregular sampling by the method of network sampling. After summarizing and comparing the common datasets and evaluation indicators for dynamic network link prediction, we briefly review classic related algorithms in recent years, and classify them according to the network changes, sampling methods, underlying principles of algorithms, and other classification methods. Meanwhile, the basic ideas, advantages, and disadvantages of these algorithms are discussed in detail. The application fields and challenges in this area are also summarized. In the final summary of the paper, the future research directions such as link prediction in dynamic heterogeneous weighted networks and the security issues brought about by link prediction are discussed.
- 2 December 2023
- Published by Springer Nature in Social Network Analysis and Mining
- Vol. 14 (1) , 1-11
VGCas: distinguishing the cascade structure and the global structure in popularity prediction
- 2 December 2023
- Published by Springer Nature in Social Network Analysis and Mining
- Vol. 14 (1) , 1-11
VGCas: distinguishing the cascade structure and the global structure in popularity prediction
- 8 July 2022
- Published by Springer Nature
- p. 494-508
Link Prediction with Mixed Structure Attribute of Network
- 8 July 2022
- Published by Springer Nature
- p. 494-508
Link Prediction with Mixed Structure Attribute of Network