Sequential recommendation has emerged as a fundamental task in various domains, aiming to predict a user's next interaction based on historical behavior. Recent advances in deep sequence models, particularly Transformer-based architectures and the more recent Mamba, have substantially pushed the boundaries of sequential modeling performance. However, existing methods still face two critical challenges. First, many current approaches overlook the hierarchical structures and high-order dependencies among items, typically restricting representation learning to conventional Euclidean spaces, which limits their capacity to capture complex relational information. Second, although Mamba excels at long-range dependency modeling, its reliance on static Feed-Forward Networks (FFNs) hinders its ability to dynamically adapt to evolving user preferences across diverse contexts. To address these limitations, we propose a Hyperbolic-Enhanced Mixture-of-Experts Mamba recommender (HM2Rec) for sequential recommendation. HM2Rec first encodes user-item relationships through hyperbolic graph convolution to exploit hierarchical structure more effectively. Then, a Variational Graph Auto-Encoder (VGAE) is employed to reconstruct node embeddings, improving structural robustness. To further enhance sequential modeling, we integrate Rotary Positional Encoding (RoPE) into Mamba to better capture relative position dependencies, and replace the FFN with Mixture-of-Expert (MOE) module, enabling dynamic and personalized expert selection for each token. Our extensive experiments on four widely-used public datasets demonstrate that HM2Rec outperforms several advanced baseline models.

Recommender Systems (RSs) are widely applied for navigating information, and Collaborative Filtering (CF) is one of prominent recommendation techniques due to the advantages of domain independence and easy interpretation. Among the numerous CF methods, Variational Autoencoders (VAE), benefiting from modeling in a probabilitistic way, stands out in capturing user preferences through representation learning. Despite the superiority, VAE-based CF models still suffer from two challenging problems: (1) Exposure bias: models in training state are narrowly exposed to a limited, biased sample of data, leading to a skewed understanding of users' true preferences; (2) Posterior collapse: models excessively simplify the learned latent variable distributions, generating na"ive representations that are unable to encapsulate the complex data patterns and thereby resulting improper recommendations. In this paper, we propose a Debiased and Representation-enhanced Variational AutoEncoder (DR-VAE) framework for collaborative recommendations. Specifically, for exposure bias problem, DR-VAE incorporates a Debiasing Estimator, mitigating the impact of exposure bias. For poster collapse issue, DR-VAE innovatively introduces a Flow-based Representation Enhancement module, ensuring us to encapsulate complex data patterns by fitting complex and intricate posterior distributions directly. We provide experimental validations over four datasets to substantiate the efficacy of our DR-VAE framework.

Time series forecasting has significant applications across various domains, including industry, agriculture, and finance. Transformer-based models have shown significant promise in enhancing time series forecasting over the past few years. However, existing methods struggle to simultaneously capture local details and global semantics under single-view architectures. They also find it difficult to dynamically adapt to time-varying and multi-scale temporal patterns while accurately modeling the complex, time-varying relationships between multiple variables. To address these challenges, we propose HRCformer, a novel Transformer-based framework that introduces two key innovations: the Hierarchical Recursive Interaction Convolution (HRIC) and the Triad Adaptive Recalibration Module (TARM). HRIC achieves joint modeling of fine-grained short-term fluctuations and high-order cross-period dependencies in time series by integrating Divide-and-Process Convolution for local processing with Recursive Channel Interaction Convolution for global processing. TARM further enhances dynamic modeling via Dynamic Variance Attention, which amplifies critical temporal deviations through 3D attention, and the Adaptive Multivariate Recalibration, which uses a two-layer fully connected network with nonlinear activation to learn the dynamic relationships between channels, suppresses noise, and emphasizes informative multivariate interactions. Comprehensive experiments conducted on seven real-world datasets highlight the superiority of HRCformer compared to prior state-of-the-art methods.

Recommender systems play a critical role in many applications by providing personalized recommendations based on user interactions. However, it remains a major challenge to capture complex sequential patterns and address noise in user interaction data. While advanced neural networks have enhanced sequential recommendation by modeling high-order item dependencies, they typically assume that the noisy interaction data as the user's preferred preferences. This assumption can lead to suboptimal recommendation results. We propose a Variational Graph Auto-Encoder driven Graph Enhancement (VGAE-GE) method for robust augmentation in sequential recommendation. Specifically, our method first constructs an item transition graph to capture higher-order interactions and employs a Variational Graph Auto-Encoder (VGAE) to generate latent variable distributions. By utilizing these latent variable distributions for graph reconstruction, we can improve the item representation. Next, we use a Graph Convolutional Network (GCN) to transform these latent variables into embeddings and infer more robust user representations from the updated item embeddings. Finally, we obtain the reconstructed user check-in data, and then use a Mamba-based recommender to make the recommendation process more efficient and the recommendation results more accurate. Extensive experiments on five public datasets demonstrate that our VGAE-GE model improves recommendation performance and robustness.

Test-time adaptation (TTA) offers the potential to enhance model generalizability without relying on training data or retraining processes. However, TTA faces challenges under covariate shift, where discrepancies between the distributions of training and testing phases hinder model performance. This limitation stems from the fact that existing methods usually rely heavily on training data and fail to establish a good connection between the model and the marginal distribution of test data, resulting in reduced generalization ability. To mitigate this issue, we introduce a novel self-supervised framework that integrates latent score matching and pseudo-label refinement into the TTA paradigm to enhance the model's perception of the test data distribution. Our approach, Joint Test-time Adaptation with Refined Pseudo-labels and Latent Score Matching, reinterprets a classifier as a score estimator and trains it using pseudo-label refinement. This enables the model to better align with the test distribution through latent score matching, while simultaneously preserving discriminative performance via pseudo-label refinement. Extensive experiments across diverse architectures and benchmarks demonstrate that TAPS consistently outperforms state-of-the-art methods in terms of generalization performance under various distribution shifts.

Next Point-of-Interest (POI) recommendation has become a crucial task in Location-Based Social Networks (LBSNs), which provide personalized recommendations by predicting the user's next check-in locations. Commonly used models including Recurrent Neural Networks (RNNs) and Graph Convolutional Networks (GCNs) have been widely explored. However, these models face significant challenges, including the difficulty of capturing the hierarchical and tree-like structure of POIs in Euclidean space and the sparsity problem inherent in POI recommendations. To address these challenges, we propose a Hyperbolic Variational Graph Auto-Encoder (HVGAE) for next POI recommendation. Specifically, we utilize a Hyperbolic Graph Convolutional Network (Hyperbolic GCN) to model hierarchical structures and tree-like relationships by converting node embeddings from euclidean space to hyperbolic space. Then we use Variational Graph Auto-Encoder (VGAE) to convert node embeddings to probabilistic distributions, enhancing the capture of deeper latent features and providing a more robust model structure. Furthermore, we combine the Mamba4Rec recommender and Rotary Position Embedding (RoPE) and propose Rotary Position Mamba (RPMamba) to effectively utilize POI embeddings rich in sequential information, which improves the accuracy of the next POI recommendation. Extensive experiments on three public datasets demonstrate the superior performance of the HVGAE model.

As an advanced carrier of on-board sensors, connected autonomous vehicle (CAV) can be viewed as an aggregation of self-adaptive systems with monitor-analyze-plan-execute (MAPE) for vehicle-related services. Meanwhile, machine learning (ML) has been applied to enhance analysis and plan functions of MAPE so that self-adaptive systems have optimal adaption to changing conditions. However, most of ML-based approaches don’t utilize CAVs’ connectivity to collaboratively generate an optimal learner for MAPE, because of sensor data threatened by gradient leakage attack (GLA). In this article, we first design an intelligent architecture for MAPE-based self-adaptive systems on web 3.0-based CAVs, in which a collaborative machine learner supports the capabilities of managing systems. Then, we observe by practical experiments that importance sampling of sparse vector technique (SVT) approaches cannot defend GLA well. Next, we propose a fine-grained SVT approach to secure the learner in MAPE-based self-adaptive systems that uses layer and gradient sampling to select uniform and important gradients. At last, extensive experiments show that our private learner spends a slight utility cost for MAPE (e.g., \(0.77\%\) decrease in accuracy) defending GLA and outperforms the typical SVT approaches in terms of defense (increased by \(10\) – \(14\%\) attack success rate) and utility (decreased by \(1.29\%\) accuracy loss).

Recently, patch-based transformer methods have demonstrated strong effectiveness in time series forecasting. However, the complexity of self-attention imposes demands on memory and compute resources. In addition, though patches can capture comprehensive temporal information while preserving locality, temporal information within patches remains important for time series prediction. The existing methods mainly focus on modeling long-term dependencies across patches, while paying little attention to the short-term dependencies within patches. In this paper, we propose the Global and Local Frequency-domain Network (GLFNet), a novel architecture that efficiently learns global time dependencies and local time relationships in the frequency domain. Specifically, we design a frequency filtering layer to learn the temporal interactions instead of self-attention. Then we devise a dual filtering block consisting of global filter block and local filter block which learns the global dependencies across patches and local dependencies within patches. Experiments on seven benchmark datasets demonstrate that our approach achieve superior performance with improved efficiency.

Federated learning addresses privacy concerns in multimedia recommender systems by enabling collaborative model training without exchanging raw data. However, existing federated recommendation models are mainly based on basic backbones like Matrix Factorization (MF), which are inadequate to capture complex implicit interactions between users and multimedia content. Graph Convolutional Networks (GCNs) offer a promising method by utilizing the information from high-order neighbors, but face challenges in federated settings due to problems such as over-smoothing, data heterogeneity, and elevated communication expenses. To resolve these problems, we propose a Cluster-driven Personalized Federated Recommender System with Interest-aware Graph Convolution Network (CPF-GCN) for multimedia recommendation. CPF-GCN comprises a local interest-aware GCN module that optimizes node representations through subgraph-enhanced adaptive graph convolution operations, mitigating the over-smoothing problem by adaptively extracting information from layers and selectively utilizing high-order connectivity based on user interests. Simultaneously, a cluster-driven aggregation approach at the server significantly reduces communication costs by selectively aggregating models from clusters. The aggregation produces a global model and cluster-level models, combining them with the user's local model allows us to tailor the recommendation model for the user, achieving personalized recommendations. Moreover, we propose an adversarial optimization technique to further augment the robustness of CPF-GCN. Experiments on three datasets demonstrate that CPF-GCN significantly outperforms the state-of-the-art models.

In many fields, such as social networks and recommendation systems with high time requirements, fake news and false information are often released in real time, impacting on people’s daily life. Entity alignment (EA) in temporal knowledge graph (TKG) can fuse the information contained in entities by finding equivalent entities, thus helping to determine the regular pattern of disinformation under time change. The existing methods either ignore the use of temporal attributes’ information and structural information or the modeling of that is insufficient, which has become a major obstacle to the further and wider application of TKG EA. In this article, we put forward a new idea of training for the processing of time attributes and relational structure information, to further enhance the ability in the EA process of TKGs. By forming box embedding matrix and name embedding matrix, and adaptively fusing the above information, we propose a new TKG EA solution. We carry out comparative experiments on standard news media and social media datasets collected from the real world, which validates the effectiveness of our proposal.

The provision of privacy-preserving recommendations for geological tourist attractions is an important research area. The historical check-in data collected from location-based social networks (LBSNs) can be utilized to mine their preferences, thereby facilitating the promotion of the geological tourism industry. However, such check-ins often contain sensitive user information that poses privacy leakage risks. To address this issue, some methods have been proposed to develop privacy-preserving point-of-interest (POI) recommendation systems. These methods commonly rely on either perturbation-based or federated learning techniques to protect users’ privacy. However, the former can hinder preference capture, while the latter remains vulnerable to privacy breaches during the parameter-sharing process. To overcome these challenges, we propose a novel privacy-preserving POI recommendation model that incorporates users’ privacy preferences based on a simplified graph convolutional neural network. Specifically, we employ a generative model to create a subset of POIs that reflect users’ preferences but do not reveal their private information, and then we design a simplified graph convolutional network to analyze the high-order connectivity between users and POIs that are privacy-preserving. The resulting model enables efficient POI recommendation under strict privacy protection, which is particularly relevant to geological tourism. Experimental results on two public datasets demonstrate the effectiveness of our proposed approach.

With the construction and development of smart cities, accurate and real-time traffic prediction plays a vital role in urban traffic. However, traffic data has the characteristics of nonlinearity, nonstationary, and complex structure, so traffic prediction has always been a challenging problem. The traditional statistical model is good at dealing with linear data and poor at dealing with nonlinear data. Although the ability to capture nonlinear data has improved, the deep learning approach has difficulty in meeting the real-time requirements of traffic prediction. To solve the above challenges, we propose a novel approach based on the autoregressive integrated moving average model (ARIMA) model and combining empirical mode decomposition (EMD) and singular value decomposition (SVD) technology, i.e., ESARIMA. This method first uses EMD to stabilize the traffic data, then uses SVD to compress data and reduce the noise, so as to improve the efficiency and accuracy of ARIMA model in predicting traffic flow. Finally, we use real data sets to verify the feasibility of ESARIMA. The experimental results show that our method outperforms state-of-the-art baselines.

Extensive user check-in data incorporating user preferences for location is collected through Internet of Things (IoT) devices, including cell phones and other sensing devices in location-based social network. It can help traveling enterprises intelligently predict users' interests and preferences, provide them with scientific tourism paths, and increase the enterprises income. Thus, successive point-of-interest (POI) recommendation has become a hot research topic in augmented Intelligence of Things (AIoT). Presently, various methods have been applied to successive POI recommendations. Among them, the recurrent neural network-based approaches are committed to mining the sequence relationship between POIs, but ignore the high-order relationship between users and POIs. The graph neural network-based methods can capture the high-order connectivity, but it does not take the dynamic timeliness of POIs into account. Therefore, we propose anInteraction-enhanced andTime-awareGraphConvolutionNetwork (ITGCN) for successive POI recommendation. Specifically, we design an improved graph convolution network for learning the dynamic representation of users and POIs. We also designed a self-attention aggregator to embed high-order connectivity into the node representation selectively. The enterprise management systems can predict the preferences of users, which is helpful for future planning and development. Finally, experimental results prove that ITGCN brings better results compared to the existing methods.

With the continuous increment of city volume and size, a number of traffic-related urban units (e.g., vehicles, roads, buildings, etc.) are emerging rapidly, which plays a heavy burden on the scientific traffic control of smart cities. In this situation, it is becoming a necessity to utilize the sensor data from massive cameras deployed at city crossings for accurate traffic flow prediction. However, the traffic sensor data are often distributed and stored by different organizations or parties with zero trust, which impedes the multi-party sensor data sharing significantly due to privacy concerns. Therefore, it requires challenging efforts to balance the trade-off between data sharing and data privacy to enable cross-organization traffic data fusion and prediction. In light of this challenge, we put forward an accurate LSH (locality-sensitive hashing)-based traffic flow prediction approach with the ability to protect privacy. Finally, through a series of experiments deployed on a real-world traffic dataset, we demonstrate the feasibility of our proposal in terms of prediction accuracy and efficiency while guaranteeing sensor data privacy.

The Corona Virus Disease 2019 has a great impact on public health and public psychology. People stay at home for a long time and rarely go out. With the improvement of the epidemic situation, people began to go to different places to check in. To maintain public mental health, it is necessary to propose a point-of-interest (POI) prediction model which can mine users' interests. However, the current techniques suffer from lower precision during prediction and the practical value is poor, which is due to the sparse data of users' check-in. Faced with this challenge, we propose an attention-based bidirectional gated recurrent unit (GRU) model for POI category prediction (ABG_poic). We regard the user's POI category as the user's interest preference because the fuzzy POI category is easier to reflect the user's interest than the POI. This method can alleviate the data sparsity, and protect users' location privacy. Since users' preferences are variable, we utilize a bidirectional GRU to capture the dynamic dependence of users' check-ins. Furthermore, since the neural network is similar to a “black box” in feature learning, the decision-making stage is opaque. Thus, we combine the attention mechanism with bidirectional GRU to selectively focus on historical check-in records, which can improve the interpretability of the model. Considering the time impact on users' check-in, we utilize the time sliding window in the ABG_poic model. Experiments on two data sets demonstrate that our ABG_poic outperforms the comparison models for POI category prediction on sparse check-in data.

In location-based social networks (LBSNs), extensive user check-in data incorporating user preferences for location is collected through Internet of Things devices, including cell phones and other sensing devices. However, directly acquiring the preferences of spars users remains an open challenge. This article offers a point-of-interest (POI) category recommendation model based on group preferences (PPCM). This model is proposed for three reasons: 1) because data influence the training of a deep learning model, the group influence of users is taken into account. To protect the privacy of users’ check-in records and classify similar users into the same group, locality-sensitive hashing (LSH) is used; 2) a successive POI category recommendation model should capture the long- and short-term dependence ability. The attention mechanism and a temporal sliding window are paired with the long short-term memory (LSTM). This paradigm is useful for efficiently mining users’ long-term dependencies and interests; and 3) although the overall users’ check-in data are vast, check-in location options are also massive for a single user. There is a scarcity of data that may be utilized to mine user interests. Thus, instead of using POI, we leverage the POI category to better mine the user’s interests. On real check-in data sets from New York City and Tokyo, the PPCM is compared to other models. The comparison results indicate that the PPCM has improved recommendation performance.

With the continuous accumulation of users' check-in data, we can gradually capture users' behavior patterns and mine users' preferences. Based on this, the next point-of-interest (POI) recommendation has attracted considerable attention. Its main purpose is to simulate users' behavior habits of check-in behavior. Then, different types of context information are used to construct a personalized recommendation model. However, the users' check-in data are extremely sparse, which leads to low performance in personalized model training using recurrent neural network. Therefore, we propose a category-aware gated recurrent unit (GRU) model to mitigate the negative impact of sparse check-in data, capture long-range dependence between user check-ins and get better recommendation results of POI category. We combine the spatiotemporal information of check-in data and take the POI category as users' preference to train the model. Also, we develop an attention-based category-aware GRU (ATCA-GRU) model for the next POI category recommendation. The ATCA-GRU model can selectively utilize the attention mechanism to pay attention to the relevant historical check-in trajectories in the check-in sequence. We evaluate ATCA-GRU using a real-world data set, named Foursquare. The experimental results indicate that our ATCA-GRU model outperforms the existing similar methods for next POI recommendation.

With the development of mobile Internet, more and more individuals and institutions tend to express their views on certain things (such as software and music) on social platforms. In some online social network services, users are allowed to label users with similar interests as “trust” to get the information they want and use “distrust” to label users with opposite interests to avoid browsing content they do not want to see. The networks containing such trust relationships and distrust relationships are named signed social networks (SSNs), and some real-world complex systems can be also modeled with signed networks. However, the sparse social relationships seriously hinder the expansion of users’ social circle in social networks. In order to solve this problem, researchers have done a lot of research on link prediction. Although these studies have been proved to be effective in the unsigned social network, the prediction of trust and distrust in SSN has not achieved good results. In addition, the existing link prediction research does not consider the needs of user privacy protection, so most of them do not add privacy protection measures. To solve these problems, we propose a trust-based missing link prediction method (TMLP). First, we use the simhash method to create a hash index for each user. Then, we calculate the Hamming distance between the two users to determine whether they can establish a new social relationship. Finally, we use the fuzzy computing model to determine the type of their new social relationship (e.g., trust or distrust). In the paper, we gradually explain our method through a case study and prove our method’s feasibility.