Fake news detection on social media is crucial to purifying the online environment and protecting public safety. Many existing methods explore the news propagation structures through graph neural networks (GNNs) to determine the truthfulness of news. End-to-end supervised GNNs notoriously depend on large amounts of labels. Recently, self-supervised graph pretraining has been a promising solution to alleviate the dependence on labels. However, the application of graph pretraining in fake news detection still suffers from two challenges: 1) the missing and unreliable interactions intrinsic in the news propagation structures seriously damage the pretraining performance. 2) There is an inherent gap between pretraining and downstream fake news detection tasks due to inconsistency in optimization objectives, which hinders the efficient transfer of pretrained prior knowledge and causes suboptimal detection results. To address the above two challenges, we propose RGCP, a structure redefined graph pretraining with contrastive prompting for fake news detection. Specifically, we design a propagation structure refinement module that adds potential implicit interactions and removes noisy interactions according to the connection probabilities between posts estimated under the guidance of self-supervised contrastive learning. Thereby, the redefined structures provide reliable news propagation patterns to generate robust pretrained news representations. Moreover, we propose a novel prompt tuning based on the contrastive learning module that reformulates the downstream fake news detection task in a similar form as the graph contrastive pretraining, bridging the optimization objective gap. The extensive experiments on benchmark datasets demonstrate the superiority of RGCP, achieving an average improvement of 10.15% in few-shot classification.

Building accurate human behavior models is necessary for ambient intelligence. However, human activity recognition (HAR) in continuously monitored physical space faces challenges to achieve a good performance yet using only simple computing resources. In this work, we model HAR as an edge classification problem for a collaborative event graph of context entities in a sequential bipartite graph form. We design a semantic learning framework, called KGAR, to perform HAR by mining, encoding, and exploiting deep semantic knowledge of activities in an end-to-end fashion. KGAR has three components: preprocessor, KGEncoder, and predictor. The preprocessor builds offline a tiny knowledge graph of activities, to model and capture multidimensional semantic relationships between activities and core context entities. KGEncoder encodes the knowledge graph of activities using improved graph neural networks (GNNs) models, to avoid different confusing context patterns. The predictor can be deployed using lightweight deep neural networks to produce real-time labels. Experimental results show that using KGEncoder in KGAR improves the performance of original deep neural networks by 25% - 439% on five datasets. The time of labeling each sensor event during testing with event streams is less than 0.5ms. We have also conducted extensive experimental study to show that KGAR outperforms different types of models in more complex activity scenarios. We believe KGAR could be used for real-time HAR in real life with its high prediction performance and low computing requirement.

Casting sequential recommendation (SR) as a reinforcement learning (RL) problem is promising and some RL-based methods have been proposed for SR. However, these models are sub-optimal due to the following limitations: (a) they fail to leverage the supervision signals in the RL training to capture users’ explicit preferences, leading to slow convergence; and (b) they do not utilize auxiliary information (e.g., knowledge graph) to avoid blindness when exploring users’ potential interests. To address the above-mentioned limitations, we propose a multiplex information-guided RL model (MELOD), which employs a novel RL training framework with Teach and Explore components for SR. We adopt a Teach component to accurately capture users’ explicit preferences and speed up RL convergence. Meanwhile, we design a dynamic intent induction network (DIIN) as a policy function to generate diverse predictions. We utilize the DIIN for the Explore component to mine users’ potential interests by conducting a sequential and knowledge information joint-guided exploration. Moreover, a sequential and knowledge-aware reward function is designed to achieve stable RL training. These components significantly improve MELOD’s performance and convergence against existing RL algorithms to achieve effectiveness and efficiency. Experimental results on seven real-world datasets show that our model significantly outperforms state-of-the-art methods.

Explicit and implicit knowledge about users and items have been used to describe complex and heterogeneous side information for recommender systems (RSs). Many existing methods use knowledge graph embedding (KGE) to learn the representation of a user-item knowledge graph (KG) in low-dimensional space. In this article, we propose a lightweight end-to-end joint learning framework for fusing the tasks of KGE and RSs at the model level. Our method proposes a lightweight KG embedding method by using bidirectional bijection relation-type modeling to enable scalability for large graphs while using self-adaptive negative sampling to optimize negative sample generating. Our method further generates the integrated views for users and items based on relation-types to explicitly model users’ preferences and items’ features, respectively. Finally, we add virtual “recommendation” relations between the integrated views of users and items to model the preferences of users on items, seamlessly integrating RS with user-item KG over a unified graph. Experimental results on multiple datasets and benchmarks show that our method can achieve a better accuracy of recommendation compared with existing state-of-the-art methods. Complexity and runtime analysis suggests that our method can gain a lower time and space complexity than most of existing methods and improve scalability.

Real-time/online activity recognition (AR) is an important technology in smart Internet of Things (IoT) systems where users are assisted by smart devices in their daily activities. How to generate appropriate feature representation from sensor event streaming is a challenging issue for accurate and efficient real-time AR. Previous AR models that rely on explicit domain knowledge are not appropriate for online recognition of complex human activities. We propose to use unsupervised learning to learn about the latent knowledge and embed the activity probability distribution prediction as high-level features to boost real-time AR performance. The proposed approach first learns the latent knowledge from explicit-activity window sequences using unsupervised learning, and derives the probability distribution prediction of activity classes for a given sliding window. Our approach then feeds the prediction with other basic features of the sliding window into a classifier to produce the final class result on each event-count sliding window. Experiments on five smart home datasets show that the proposed method achieves a higher accuracy by at least 20 percent improvement on F1_score than previous traditional algorithms, while maintaining a lower time cost than deep learning based methods. An analysis on the feature importance shows that the addition of probability distribution prediction about activity classes leads to a promising direction for real-time AR.

Social recommender system, using social relation networks as additional input to improve the accuracy of traditional recommender systems, has become an important research topic. However, most existing methods utilize the entire user relationship network with no consideration to its huge size, sparsity, imbalance, and noise issues. This may degrade the efficiency and accuracy of social recommender systems. This study proposes a new approach to manage the complexity of adding social relation networks to recommender systems. Our method first generates an individual relationship network (IRN) for each user and item by developing a novel fitting algorithm of relationship networks to control the relationship propagation and contracting. We then fuse matrix factorization with social regularization and the neighborhood model using IRN's to generate recommendations. Our approach is quite general, and can also be applied to the item-item relationship network by switching the roles of users and items. Experiments on four datasets with different sizes, sparsity levels, and relationship types show that our approach can improve predictive accuracy and gain a better scalability compared with state-of-the-art social recommendation methods.

Improving quality of services (QoS) through applying trust and reputation management technology is increasingly popular in the literature and industry. Most existing trust and reputation systems calculate a general trust value or vector based on the gathered feedback without regard to trust's locality and subjectivity; therefore, they cannot effectively support a personal selection with consumer preferences. Our goal is to build a trust and reputation mechanism for facilitating a trustworthy and personal service selection in a service-oriented Web, where service peers can act as a service provider and/or a service consumer. A user-centric trust and reputation mechanism distinguishing the different trust context and content to enable a personal service selection with regard to trust preference in a service-oriented Web is represented in detail. It is widely recognized that reputation-based trust methods must face the challenge of malicious behaviors. To deal with the malicious feedback behaviors, we introduce a “bidirectional'' feedback mechanism based on QoS experience similarity in our trust and reputation framework. The test run demonstrates that our method can significantly increase the success rate of service transactions and is effective in resisting various malicious behaviors of service peers, when it is compared to other similar methods. © 2011 Wiley Periodicals, Inc.

In open and dynamic environments, reputation-based trust is a key issue for service selection which is based on the prediction of services' performance. Typically, the trust and reputation evaluation is based on the feedback of the service quality from all service users. The crucial issue in this evaluation is to detect and deal with false or unfair feedback from dishonest providers and cheating users. In this paper, we first present a dynamic service selection framework then introduce a semantic match and rank algorithm to enable the automatic selection. Furthermore, we propose a reputation assessment approach to address trust and reputation of service and service user in various realistic cheating behaviors. Finally, to the best of our knowledge, a set of empirical studies have been conducted to demonstrate that our solution gains good results under collaboration attacks of various cheating behaviors.