Funing Yang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Funing Yang has a total of 20 co-authors. Here are the top co-authors by number of shared publications:
- Yongjian Yang
College of Computer Science and Technology, Jilin University, Changchun, China
3 shared publications - En Wang
Jilin University
2 shared publications - Hepeng Gao
Jilin University, Changchun, Jilin, China
2 shared publications - Wenbin Liu
Jilin University
2 shared publications - Xingliang Zhang
China Mobile Group Jilin Co., Ltd, China
2 shared publications - Bohui Du
College of Computer Science and Technology, Jilin University, Changchun, Jilin, China
1 shared publication
Latest Publications
- 29 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. PP (15455971) , 1-16
Privacy-Preserving Data Completion With Location and Value Obfuscation in Spatial Crowdsourcing
Spatial-temporal data is essential for various applications. Traditional data collection methods often have limited coverage, while Spatial Crowdsourcing, which leverages mobile devices, offers broader data acquisition but also presents challenges with data sparsity. To infer missing data based on observed data, existing data completion techniques often require detailed spatial-temporal information, raising privacy concerns. Some methods attempt to protect privacy by obfuscating locations; however, this can disrupt the original spatial-temporal correlations, leading to increased errors in data completion. To address this, we propose a differential privacy-based obfuscation strategy that safeguards privacy while maintaining data accuracy. Our approach employs a prediction model to extract an importance matrix and a relationship matrix from historical data. These matrices are utilized to map actual user locations to more beneficial locations for data completion, followed by value adjustments to maintain availability and preserve spatial-temporal correlations. We demonstrate that our mechanism satisfies differential privacy that provides effective privacy guarantees. Evaluations of five real-world datasets show that our obfuscation mechanism outperforms other baseline methods at the same level of privacy protection.
- 29 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. PP (15455971) , 1-16
Privacy-Preserving Data Completion With Location and Value Obfuscation in Spatial Crowdsourcing
Spatial-temporal data is essential for various applications. Traditional data collection methods often have limited coverage, while Spatial Crowdsourcing, which leverages mobile devices, offers broader data acquisition but also presents challenges with data sparsity. To infer missing data based on observed data, existing data completion techniques often require detailed spatial-temporal information, raising privacy concerns. Some methods attempt to protect privacy by obfuscating locations; however, this can disrupt the original spatial-temporal correlations, leading to increased errors in data completion. To address this, we propose a differential privacy-based obfuscation strategy that safeguards privacy while maintaining data accuracy. Our approach employs a prediction model to extract an importance matrix and a relationship matrix from historical data. These matrices are utilized to map actual user locations to more beneficial locations for data completion, followed by value adjustments to maintain availability and preserve spatial-temporal correlations. We demonstrate that our mechanism satisfies differential privacy that provides effective privacy guarantees. Evaluations of five real-world datasets show that our obfuscation mechanism outperforms other baseline methods at the same level of privacy protection.
- 9 May 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 327
Temporal-aware dynamic graph neural networks for next POI recommendation
- 9 May 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 327
Temporal-aware dynamic graph neural networks for next POI recommendation
- Open Access
- 22 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 3377-3388
Fine-Grained Data Inference via Incomplete Multi-Granularity Data
- 22 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 3377-3388
Open AccessFine-Grained Data Inference via Incomplete Multi-Granularity Data
- 12 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (6) , 7293-7305
Future-Aware Balanced Preference Matching for Real-Time On-Demand Taxi Dispatch
Spatial crowdsourcing is drawing much attention with the rapid development of mobile Internet. Achieving efficient crowdsourcing task assignment involves not only maximizing the earnings of workers but also balancing the preferences of users or customers. Users often express preferences for specific workers or conditions, such as particular drivers, delivery personnel, or service providers. To address this challenge, we investigate the future-aware balanced preference (FABP) problem. This problem aims to maximize the profits of global workers while simultaneously considering the preferences of both parties to ensure bilateral satisfaction. To address the FABP problem, we propose the learning to match (LTM) algorithm. This algorithm utilizes online reinforcement learning that considers both immediate profits and long-term rewards. It acknowledges the significance of task assignment decisions in relation to the spatial distribution of future drivers, which in turn affects subsequent decisions. The LTM algorithm generates future-aware preference lists using learned driver state values and guides the subsequent matching. Additionally, we present the real-time preference-based matching (RTPM) algorithm, which is a real-time matching algorithm that enables substitutions based on preference lists when a more preferred matching pair becomes available. This enhances the efficiency and fairness of real-time task assignment in dynamic environments, while simultaneously meeting the needs of passengers and drivers. Our extensive experiments on both real and synthetic datasets validate the effectiveness of our proposed algorithms, demonstrating a noteworthy improvement of up to 11.8% and an average increase of 4.7% compared to benchmark algorithms.
- 12 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (6) , 7293-7305
Future-Aware Balanced Preference Matching for Real-Time On-Demand Taxi Dispatch
Spatial crowdsourcing is drawing much attention with the rapid development of mobile Internet. Achieving efficient crowdsourcing task assignment involves not only maximizing the earnings of workers but also balancing the preferences of users or customers. Users often express preferences for specific workers or conditions, such as particular drivers, delivery personnel, or service providers. To address this challenge, we investigate the future-aware balanced preference (FABP) problem. This problem aims to maximize the profits of global workers while simultaneously considering the preferences of both parties to ensure bilateral satisfaction. To address the FABP problem, we propose the learning to match (LTM) algorithm. This algorithm utilizes online reinforcement learning that considers both immediate profits and long-term rewards. It acknowledges the significance of task assignment decisions in relation to the spatial distribution of future drivers, which in turn affects subsequent decisions. The LTM algorithm generates future-aware preference lists using learned driver state values and guides the subsequent matching. Additionally, we present the real-time preference-based matching (RTPM) algorithm, which is a real-time matching algorithm that enables substitutions based on preference lists when a more preferred matching pair becomes available. This enhances the efficiency and fairness of real-time task assignment in dynamic environments, while simultaneously meeting the needs of passengers and drivers. Our extensive experiments on both real and synthetic datasets validate the effectiveness of our proposed algorithms, demonstrating a noteworthy improvement of up to 11.8% and an average increase of 4.7% compared to benchmark algorithms.
- Open Access
- 26 July 2024
- Published by Association for Computing Machinery (ACM)
- p. 154-157
Information Fusion Representation Learning in Zero-shot Scenarios
Knowledge Representation Learning (KRL), also known as Knowledge Graph Embedding, is an important field in KG research that deals with knowledge acquisition, representation, and inference. However, current KRL faces challenges in learning entity representations in Zero-Shot scenarios. To address this, the ZFRL model is introduced. It efficiently handles entity and link prediction in an open-domain knowledge graph using two components: a structured knowledge embedding layer and a novel attribute encoding layer. Experimental results on an open-domain knowledge graph dataset show significant improvements in entity and relationship prediction accuracy. The ZFRL model effectively links newly added entities in a KG, supporting KG completion tasks.
- 26 July 2024
- Published by Association for Computing Machinery (ACM)
- p. 154-157
Open AccessInformation Fusion Representation Learning in Zero-shot Scenarios
Knowledge Representation Learning (KRL), also known as Knowledge Graph Embedding, is an important field in KG research that deals with knowledge acquisition, representation, and inference. However, current KRL faces challenges in learning entity representations in Zero-Shot scenarios. To address this, the ZFRL model is introduced. It efficiently handles entity and link prediction in an open-domain knowledge graph using two components: a structured knowledge embedding layer and a novel attribute encoding layer. Experimental results on an open-domain knowledge graph dataset show significant improvements in entity and relationship prediction accuracy. The ZFRL model effectively links newly added entities in a KG, supporting KG completion tasks.
- Open Access
- 26 July 2024
- Published by Association for Computing Machinery (ACM)
- p. 150-153
Research on Information Fusion Representation Learning Method Based on BERT
Knowledge Graph (KG) is a structured knowledge representation model. However, the current field of KG has the problem of incomplete utilization of attribute information of entities, and there is no effective combination of structural information in knowledge graphs and attribute information of entities. In order to solve this problem, this paper proposes an innovative KRL method-BFRL, which can skillfully integrate knowledge graph structure information and entity attribute information. And through the BERT-based attribute encoder, the attribute type and attribute value are processed separately to remove the noise influence in the attribute data, thereby obtaining a reliable embedded representation of the attribute information. The effectiveness of the BFRL method is demonstrated through experiments on real-world data sets.
- 26 July 2024
- Published by Association for Computing Machinery (ACM)
- p. 150-153
Open AccessResearch on Information Fusion Representation Learning Method Based on BERT
Knowledge Graph (KG) is a structured knowledge representation model. However, the current field of KG has the problem of incomplete utilization of attribute information of entities, and there is no effective combination of structural information in knowledge graphs and attribute information of entities. In order to solve this problem, this paper proposes an innovative KRL method-BFRL, which can skillfully integrate knowledge graph structure information and entity attribute information. And through the BERT-based attribute encoder, the attribute type and attribute value are processed separately to remove the noise influence in the attribute data, thereby obtaining a reliable embedded representation of the attribute information. The effectiveness of the BFRL method is demonstrated through experiments on real-world data sets.