In prior research, spatial coverage has been the primary metric for assessing the quality of mobile network data, with various participant selection strategies developed for data brokers operating within budget constraints. This paper introduces a novel data participant service quality model that prioritizes participant privacy while ensuring creditworthiness and participation willingness. This paper proposes a selection strategy for mobile network data participants that operates under the principles of local differential privacy, designed to maximize the marginal gain under a cost budget constraint. Drawing upon the principle of marginal gains from economics, our strategy advocates for the selection of participants whose marginal gains exceed their associated costs, thereby optimizing the selection process based on relative marginal costs. This paper formalizes this approach as the marginal problem of data participant selection in mobile network data markets, constructing a model and providing theoretical analysis to establish its NP-hard. To address this problem, we develop the Random Response Adaptive Algorithm (RRAA), designed to select participant groups that maximize mobile network data gains. Extensive experiments on both real and simulated datasets demonstrate the algorithm's efficiency and effectiveness. Additionally, to safeguard participant privacy, we propose the Random Response Self-Adaptive Algorithm (RRSAA). Our experimental evaluation on real datasets not only validates the algorithm's privacy-preserving capabilities but also elucidates the impact of privacy preservation parameters on marginal gains.

Anomaly detection in industrial chains is crucial for preventing industrial chain risks. Unlike traditional time series data, industrial chain data possesses characteristics such as temporal evolution, heterogeneous distribution, complex association structures, and non-uniform sampling, making anomaly detection particularly challenging. Existing methods are primarily designed for static homogeneous networks and uniform time sampling scenarios, making it difficult to effectively capture abnormal evolution patterns in heterogeneous association structures within non-uniform time series. To address these challenges, this paper proposes a novel Temporal Heterogeneous Graph Learning-based Anomaly Detection framework (THeGraL-AD) that leverages node and semantic attention mechanisms to enhance neural controlled differential equations for effectively capturing industrial chain evolution patterns. Specifically, we first construct a temporal heterogeneous evolution graph that integrates static topological structures with dynamic spatiotemporal dependencies. Subsequently, we design a neural controlled differential equation module combined with a dual graph attention mechanism to extract node evolution features. An adaptive threshold strategy is then utilized to detect anomalous nodes. Experiments demonstrate that THeGraL-AD outperforms baseline methods on three datasets, achieving more accurate anomaly detection.

Knowledge graph (KG) completion, a cornerstone of semantic reasoning and AI applications, overwhelming majority of them only consider learning from binary relations. Real-world knowledge is inherently N-ary, over 61 <3 neighbors), which suffer from irrelevant information aggregation and incomplete context integration. To resolve these limitations, we propose Direction-aware Attentive Hypergraph (DAHG) network, which resolve the dual challenges through direction-aware hyperedges that preserve the direction of knowledge flows and capture implicit interaction information with global context that inject into lower degree entities. We demonstrate comparative experiments on two benchmark knowledge graph datasets (WN18NN and FB5K-237), that DAHG outperforms the other state-of-the-art completion method, thereby demonstrating its ability to express N-ary directed knowledge effectively.

With the increasing complexity of industrial chains, intelligent monitoring methods for industrial chain data have become essential for economic security and high-quality industrial development. Industrial chain data has temporal and heterogeneous characteristics, requiring intelligent systems that can forecast trends and adjust monitoring strategies in response to anomalies. Current trend prediction methods struggle to model complex heterogeneous interactions between variables, while existing adaptive monitoring approaches lack sensitivity to abnormal trends and cannot efficiently sample multiple variables synchronously. To address these challenges, we propose: (1) HGAN-TPMLP, a trend prediction algorithm combining heterogeneous graph representation learning with a Patch-based MLP structure and temporal causal convolution, improving both short-term feature sensitivity and long-range forecasting; (2) HTAAM, an adaptive monitoring method including HGAN-GRU for trend anomaly detection and EATSA for time-synchronized adaptive sampling based on anomaly scores. Experiments demonstrate that our methods significantly improve prediction accuracy on industrial chain data while reducing unnecessary data transmission and enhancing anomaly coverage, making them well-suited for complex industrial environments.

Shortest path queries are among the most fundamental operations in graph processing. Prior studies have extensively examined label-constrained queries on static graphs with fixed edge weights. However, real-world networks exhibit temporal dynamics and often contain complex heterogeneous structures. Examples include peak-hour variations in travel time and the coordination of multiple transportation modes. These time-dependent and label-dependent characteristics limit the applicability of existing methods to the Time-Dependent Label-Constrained Shortest Path (TD-LCSP) problem. To address this limitation, we propose LP-Tree, a label-partitioned tree decomposition framework that models both label constraints and temporal dependencies in a unified manner. Based on this structure, we design the Parameter-Separated (PS) Index, which decouples key parameters to improve query efficiency and reduce storage costs. The combined approach significantly lowers space consumption and achieves sub-millisecond query latency on large-scale time-dependent graphs. Extensive experiments on multiple datasets demonstrate the effectiveness and efficiency of our proposed approach.

Retrieval Augmented Generation (RAG) has emerged as a mainstream paradigm for mitigating hallucinations in large language models and supporting knowledge-intensive complex reasoning tasks. However, existing retrieval-as-context methods typically employ either one-shot retrieval or blind iteration strategies, lacking global planning and dynamic scheduling for heterogeneous knowledge. This often introduces substantial redundant information and disrupts the reasoning process. To address these issues, we propose structured memory scheduling retrieval augmented generation (SMS-RAG) framework. SMS-RAG explicitly models external knowledge as compound memories with semantic, structural, and episodic features. It introduces a structured scheduling mechanism to explicitly model and control the activation timing, sequence, and dependency relationships among different knowledge. This enables knowledge to participate in reasoning in a structured manner, rather than merely serving as passive contextual input. It consists of three modules: MemPlanner determines which memory modules to activate by generating a task-aware MemoryPlan; MemScheduler decides in what order to activate them by constructing a dependency-aware graph; and MemRouter executes how to activate by selectively routing across semantic, structural, and episodic memory layers to extract critical information for reasoning. Extensive experiments on three multi-hop QA benchmarks and a domain-specific legal QA dataset demonstrate that SMS-RAG significantly outperforms existing state-of-the-art methods in reasoning accuracy, factual coverage, and scheduling efficiency.

Text-Based Person Search (TBPS) holds unique value in real-world surveillance bridging visual perception and language understanding, yet current paradigms utilizing pre-training models often fail to transfer effectively to complex open-world scenarios. The reliance on "Passive Observation" leads to multifaceted spurious correlations and spatial semantic misalignment, causing a lack of robustness against distribution shifts. To fundamentally resolve these defects, this paper proposes ICON (Invariant Counterfactual Optimization with Neuro-symbolic priors), a framework integrating causal and topological priors. First, we introduce Rule-Guided Spatial Intervention to strictly penalize sensitivity to bounding box noise, forcibly severing location shortcuts to achieve geometric invariance. Second, Counterfactual Context Disentanglement is implemented via semantic-driven background transplantation, compelling the model to ignore background interference for environmental independence. Then, we employ Saliency-Driven Semantic Regularization with adaptive masking to resolve local saliency bias and guarantee holistic completeness. Finally, Neuro-Symbolic Topological Alignment utilizes neuro-symbolic priors to constrain feature matching, ensuring activated regions are topologically consistent with human structural logic. Experimental results demonstrate that ICON not only maintains leading performance on standard benchmarks but also exhibits exceptional robustness against occlusion, background interference, and localization noise. This approach effectively advances the field by shifting from fitting statistical co-occurrences to learning causal invariance.

The Apache Flink distributed computing system demonstrates significant advantages in iterative computing. Establishing appropriate parallelism for operators is crucial to further enhancing the efficiency of Flink's iterative jobs and optimizing cluster resource utilization. However, the encapsulated user-defined functions within operators cannot be directly analyzed, posing challenges for optimizing parallelism based on computation logic. Moreover, the diverse and complex nature of hardware resources in distributed computing clusters exacerbates this issue, making it difficult to model the relationship between optimal parallelism and computational resources. To address these challenges, this paper proposes the Dynamic Parallelism Adjustment (DAPIC) framework. The DAPIC framework introduces a mechanism for dynamically adjusting operator parallelism in batch processing, leveraging an enhanced GCN + GRU model. Additionally, it incorporates a resource-aware parallelism adjustment mechanism for stream processing, considering the availability of remaining computational resources. Experimental results show that the DAPIC framework improves the efficiency of iterative jobs in batch processing by 38.94% - 138.82% and in stream processing by 28.85% - 31.08%. In heterogeneous bandwidth environments, the advantages of the DAPIC framework become even more pronounced, achieving efficiency improvements exceeding 1.5 times. Furthermore, the framework effectively reduces TaskSlot occupancy, resulting in resource savings of 24.61% - 75.78%.

With the rapid development of spatial temporal crowdsourcing applications, the online task assignment problem has been widely studied as one of the most typical problems. It ensures efficient and accurate matching between tasks and workers. Traditional task assignment only focuses on solving the task assignment on a single platform. Recently, with the widespread application of data sharing technology, cross online task assignment has been proposed, aiming at increasing the mutual benefit through cooperations. However, existing methods do not consider data privacy protection during the cooperation process, resulting in the leakage of sensitive information such as users’ location and historical data of platform. In this paper, we propose Privacy-preserving Cooperative Online Matching problem, which protects the privacy of the users and workers on their respective platforms. We design a PCOM framework and provide theoretical proof that the framework satisfies Differential Privacy. We then propose two privacy-preserving algorithms to solve PCOM. Furthermore, to reduce the impact of location perturbation on matching results, we design a new geographical location perturbation mechanism and a cooperative platform selection algorithm. Extensive experiments on real and synthetic datasets confirm the effectiveness and efficiency of our algorithms.

Traffic flow prediction is crucial for efficient urban planning, traffic management, and user navigation. Modern deep learning models have achieved great success in capturing the complex spatio-temporal dependencies in traffic networks. However, due to frequently changing traffic patterns, the performance of deployed models degrades over time, necessitating periodic updates. Full-scale model retraining is computationally expensive, creating a critical conflict between maintaining prediction accuracy and minimizing update overhead. To address this, incremental update or continual learning methods have emerged, but existing approaches are often tightly coupled with specific model architectures and rely on unclear criteria for data selection, thereby causing redundant data selection and lacking interpretability. To overcome these limitations, we propose a lightweight continual learning method based on B-splines. This method identifies the most valuable data for model updates by analyzing the intrinsic geometric and statistical properties of the traffic data itself. Specifically, we fit a B-spline curve to create a smooth representation of the core traffic pattern and then compute the regression leverage score for each data point to quantify its structural importance. This strategy decouples the data evaluation process from the internal mechanisms of the prediction model. Since the selected data points directly reflect key features of the traffic pattern-such as peaks, inflection points, and anomalies-our method is inherently interpretable, allowing users to understand why certain data points are chosen. Extensive experiments on multiple real-world datasets demonstrate that our method maintains a high level of prediction accuracy while significantly reducing the computational cost of model updates, offering an efficient and transparent solution for the maintenance of dynamic traffic systems.

As industrial chain data grows more complex, existing models struggle to capture temporal dependencies. Dynamic data is characterized by multi-sourcing and continuous evolution, while knowledge graphs enhance risk prediction by uncovering complex relationships. However, current methods treat static knowledge and temporal data separately, neglecting their interaction and failing to address evolving, long-term, and short-term patterns. To overcome this, we propose KGR-HATA, a Knowledge Graph Reasoning with Hierarchical Attention-based Temporal Aggregation method. It includes a subgraph mapping evolution module for dynamic interactions, a knowledge graph evolution module for temporal evolution, and a temporal aggregation module to capture both long- and short-term associations. Experimental results show that KGR-HATA outperforms existing methods in prediction performance.

The discovery of frequent itemset can serve valuable economic and research purposes. Mining such data, however, presents privacy challenges. Local differential privacy (LDP) has been established as a strong and rigorous privacy scheme for collecting sensitive information from users. Due to the inherent high-dimensionality and heterogeneity of set-valued data, it is quite difficult to mine frequent itemsets under LDP. Existing solutions for this issue are either limited to single items or rely on sampling-based frequency oracle protocols, leading to poor overall result quality. Motivated by this, we present PrivMiner, a two-phase LDP protocol which directly collects user record information comprehensively. While the methodology of directly collecting an entire user record is an obvious choice, obtaining an accurate record under LDP is highly challenging. PrivMiner accomplishes this with a novel technique called the Group-wise Estimation Mechanism by exploiting the similarity between records. PrivMiner needs significantly less perturbations than previous methods, and it achieves higher overall result quality, even for millions of items. Extensive experiments conducted on real-world datasets demonstrate the effectiveness of our proposed method and its advantages over existing solutions. In particular, PrivMiner improves the error of estimation count by at least one order of magnitude.

The generation and defense of text adversarial samples are crucial for improving the robustness and security of social Internet of Things (SIoT) systems, as the exchange of information between devices in SIoT relies heavily on NLP technology. However, the discrete nature of text data leads to a lack of contextual integration in current mainstream adversarial sample generation methods based on text replacement. This results in poor stealth of the generated samples and inefficiencies due to excessive queries to the target model. Meanwhile, defense methods like adversarial training are insufficiently universal and generalizable to handle the diverse and complex range of adversarial attack strategies. This article proposes a contrastive learning-based method for generating text adversarial samples and a mutual information regularization-based method for defending against text adversarial samples, tailored to the characteristics of devices in SIoT systems and the challenges mentioned above. The proposed method leverage keyword localization, optimal perturbation, and candidate set evaluation to enhance the effectiveness of adversarial samples. Additionally, by combining mutual information measures, statistical estimation functions, and idempotent constraints, the model itself is equipped with effective defenses against adversarial samples. Compared to the baseline, the proposed method significantly reduce the magnitude of perturbations and the number of access attempts to the original samples, while greatly increasing the attack success rate. When applying the proposed adversarial sample defense method, the model's accuracy showed a significant improvement after being subjected to adversarial attacks.

Spatial crowdsourcing has become popular in recent years, but traditional tasks focus on one-to-one services with single skills like food delivery and ride-hailing. As societal needs grow more complex, there is a need for tasks requiring teams with multiple skills. Current team formation methods using workers from a single platform limit skill diversity, leading to potential task delays, lower quality, and revenue losses. Although cross-platform cooperation offers a potential solution to skill diversity limitations, it faces two challenges: (1) Data protection regulations mandate that platform's raw data must remain localized; (2) cross-platform cooperation incurs additional cooperation costs. To address these challenges, we first define the Cross-platform Online Team Formation (COTF) problem. We then propose a COTF framework and Random Cooperation Strategy to solve COTF problem. To enhance the effectiveness of cooperation, we further propose Precision Query Range Optimization Strategy (PQROS) for worker selection through adaptive range queries, and Dynamic Query Optimization (DQO) for cost-effective scheduling via predictive revenue modeling. Extensive experiments on real and synthetic datasets validate the effectiveness of our proposed methods.

The Social Internet of Things (SIoT) system enables connectivity among smart devices by integrating social networks with the Internet of Things. This integration is essential for advancing intelligent services and applications, as well as enhancing the commercial value of data. Rational task and data deployment strategies allow different types of devices to perform optimally in their areas of expertise, reducing network load, improving the timeliness of data processing, and ensuring efficient collaboration across the entire system. However, the limited computational capacity and network bandwidth of SIoT devices result in communication delays that significantly impact job responsiveness and energy consumption. The heterogeneity in computational power and bandwidth across different devices leads to resource overload or underutilization with traditional data partitioning methods, further impacting the performance of SIoT systems. This article proposes an intelligent task and data deployment method to address these issues. The proposed method abstracts the job execution process as JobGraph instance model and optimizes the mapping relationship of operators within TaskProcess. Additionally, an improved linear programming model is used to optimize the data distribution ratio among operators in heterogeneous computing environments. The proposed task deployment method achieves an average improvement of 19.6%-30.2% in job efficiency while reducing interdevice data transmission by 34.2%. In heterogeneous computing environments, the combination of the two deployment optimization methods further reduces job execution time, achieving efficiency improvements of over two times in optimal scenarios.

With the proliferation of mobile smart devices and wireless communication technologies, Mobile CrowdSensing (MCS) has emerged as a significant data collection method. MCS faces two key challenges: selecting high-quality data sellers with unknown reliability and determining fair compensation that addresses device wear and privacy risks. We introduce two novel contributions. First, the MAB-RS algorithm leverages multi-armed bandit reinforcement learning and a data freshness model to dynamically optimize seller recruitment, efficiently balancing exploration of unknown sellers and exploitation of high-quality ones. Second, the MAB-RSP employs a Stackelberg game framework, enabling platforms and sellers to collaboratively maximize profits through strategic pricing and participation incentives. Experiments demonstrate that the algorithm improves revenue while ensuring balanced benefits for all participants.

Analyzing multi-channel data related to the industrial chain through graph representation learning is of significant value for industrial chain risk detection. Multi-channel data related to the industrial chain exhibits strong correlations, which enables such data to be modeled using graph structures. Furthermore, due to multi-channel data being multi-source, heterogeneous, and containing temporal information, industrial chain risk detection can be effectively conducted from the perspective of temporal heterogeneous graph representation learning. However, existing methods primarily obtain node features by sequentially processing the temporal heterogeneous information within a single graph, and they use attention based on discrete tokens to learn long-term information, which overlooks the multi-temporal information interactions between nodes and fails to effectively capture the temporal patterns in long-term information. To address these problems, we propose a multi-temporal heterogeneous graph learning with pattern-aware attention (MTHG-PA) for industrial chain risk detection. Firstly, we propose a method for interactive aggregation of multi-temporal heterogeneous information to learn node features, thereby avoiding the problem of insufficient information interaction between nodes across time. Secondly, we design a temporal pattern-aware attention module to learn the temporal patterns in long-term information, addressing the problem that traditional attention mechanisms can only handle discrete tokens. We conduct extensive experiments on three real-world datasets. The results show that our proposed method achieves superior performance in temporal heterogeneous graph learning compared to state-of-the-art methods.

In recent years, the burgeoning data market has witnessed a surge in data exchange, playing a pivotal role in augmenting the predictive and decision-making capabilities of machine learning. Despite these advancements, persistent concerns surrounding data privacy have resulted in stringent limitations on data sharing and trading. Consequently, the data market is undergoing a transformative shift from pricing individual datasets to pricing the models themselves. The challenge of training high-performance machine learning models with restricted data from a single client is substantial. Federated learning has emerged as a popular solution, allowing collaborative model training without the need to transfer client data beyond local environments. However, training federated models within a data market introduces several challenges, including the effective selection of clients for model training and the optimization of utility through model pricing. In response to these challenges, we propose the Tiered Federated Learning Client Selection Algorithm (TiFLCS-MAR), employing a multi-attribute reverse auction approach. Integrated into the federated learning framework, TiFLCS-MAR excels at the comprehensive evaluation of client attributes, employing a tiered strategy to mitigate issues arising from client heterogeneity. Additionally, we introduce the TiFLCS-MAR Pricing Framework (TiFLCS-MARP), leveraging Nash equilibrium principles to maximize profitability for both clients and servers. Our framework accommodates the heterogeneity of diverse clients, efficiently selecting suitable candidates from a large pool, thereby boosting training efficiency and curbing model pricing costs. Empirical evidence showcases the efficacy of federated training with TiFLCS-MAR, demonstrating nearly double the convergence speed and a 5-10 percentage point improvement in accuracy across real and synthetic datasets. Furthermore, when compared to three baseline algorithms, TiFLCS-MARP substantially increases central server revenue by factors of 1.99, 27.05, and 1.78, highlighting its superior performance in the data market context.

[Purposes]To address the imbalance between supply and demand in traditional sin-gle platform task assignment,Cross Online Matching(COM)has emerged as a novel solution that allows multiple similar platforms to establish cooperative relationships and send uncompleted tasks to other platforms,increasing the probability of task acceptance.However,current COM solutions only consider single-round matching processes,making it difficult to find optimal deci-sion results in multi-platform competition.To settle these limitations,the Multi-Round Cross Online Matching problem(MRCOM)is studied and Greedy-based Multi-Round Cross Online Matching(G-MRCOM)and Game-Theoretic Multi-Round Cross Online Matching(GT-MR-COM)algorithms are proposed.[Methods]G-MRCOM improves task completion efficiency by forwarding and matching tasks in multiple rounds,with platforms greedily selecting high-reward tasks to accomplish.GT-MRCOM,on the other hand,establishes incentive mechanisms among algorithms cooperating platforms,calculates task assignment strategies that satisfy Nash Equilib-rium,and enables the platform to find better strategies in competition,thereby enhancing overall performance.[Findings]Experimental results demonstrate that the proposed algorithms can in-crease the total revenue of platforms,showcasing the effectiveness and efficiency of this study.

Data have become a valuable digital resource. It has in turn precipitated the emergence of big data marketplaces. For social network date in the marketplaces, each node should be priced according to its influence. The key challenge is that deep learning based pricing models require initial cascade graphs as inputs to predict influence, which cannot be obtained while pricing nodes. Furthermore, node pricing must enhance purchase intentions while being consistent with their influence. To address these challenges, a node-pricing framework is proposed, in which market price is determined based on the predicted influence. In this framework, corrections are performed by using a graph autoencoder (GAE). The corrections are used to augment the neighborhood subgraph and facilitate the extraction of valid sequence features, which are then used to predict influence. An approximate Shapley value for node influence is used to evaluate the price of the nodes. A multi-perspective pricing approach is further investigated, where consumer utility and the approximate Shapley value for influence are the objectives. An inflection point is chosen on the Pareto frontier to select a price that enhances consumer utility. Extensive experiments were conducted on two real-world social network datasets. The results indicate that our performance is higher than DeepCas by 10.38% in Twitter and 9.64% in Weibo . The price output by our framework is consistent with the nodes’ social marketing value while maximizing consumer utility.