Microservice deployment should consider dependency constraints and resource requirements while minimizing deployment costs. There are various dependencies among microservices, and such intricate dependencies bring great difficulties to efficient deployment. Despite significant progress, existing deep learning-based methods are not expressive enough to capture the complexity of microservice dependencies. Meanwhile, metaheuristic methods encode such dependencies into problem-specific heuristics through explicit formulae, yet such formulations remain static and inherently limited. This paper addresses these issues by formulating the microservice deployment problem (MSDP) as a variable-sized bin packing problem. We first introduce a heterogeneous graph neural network to effectively represent the complex dependencies of MSDP. Based on this representation, we propose a novel reinforcement learning policy to design heuristics without the need to derive explicit formulae. Guided by the learned heuristics, we further devise a parallel ant colony optimization algorithm to accelerate solution construction. Experimental results from three representative microservice applications demonstrate that our method obtains near-optimal solutions while significantly outperforming the state-of-the-art methods in computational efficiency.

Spatial crowdsourcing platforms have become indispensable in addressing the evolving needs of modern society. These platforms facilitate essential services such as ride-sharing, on-demand food delivery, and efficient parcel distribution. However, the uneven distribution of workers and requests under a single-platform setting may lead to the loss of tasks. To address this issue, we introduce the Cross Online Matching (COM) problem, which facilitates collaboration among multiple platforms. We first propose DemCOM and RamCOM, which adopt deterministic greedy and randomized trade-off strategies, respectively. Furthermore, we develop a Utility-Distribution Aware Cooperative Online Matching (UDACOM) algorithm that leverages supply-demand relationships to optimize decision-making. Theoretical analysis confirms the competitive ratios of our algorithms. Validated on both real and synthetic datasets, our approach significantly outperforms state-of-the-art methods, achieving a 5% increase in total revenue and a 3% improvement in the successful matching rate.

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.

Social Internet of Things (SIoT) has emerged as a novel paradigm that enhances IoT service capabilities by leveraging device-level social relationships. However, the explosive growth of heterogeneous devices, the dynamic mobile device trajectories, and complex spatiotemporal interaction patterns severely hinder SIoT network navigability. Particularly, the device mobility and contextual diversity pose significant challenges to social relation classification, a critical task for efficient routing and service discovery. Existing approaches, primarily designed for static or homogeneous networks, fail to adequately capture the evolving contextual dependencies and the spatiotemporal heterogeneity in SIoT. To address these challenges, we propose Trajectory-Aware Graph LLM (TAGLLM), a novel framework that enhances SIoT navigability through context-aware relation classification. TAGLLM introduces a multi-feature fusion trajectory evolutionary graph encoder to jointly model complex device attributes, social relations, and dynamic trajectories. Furthermore, a structural graph-text token alignment strategy is designed to exploit the generalization ability and contextual understanding capabilities of Large Language Models (LLMs), enabling more effective modeling of heterogeneous and dynamic SIoT scenarios. Extensive experiments on real-world SIoT datasets demonstrate that TAGLLM outperforms state-of-the-art baselines across multiple evaluation metrics, highlighting its potential to push the frontier of graph learning and LLM integration in SIoT applications.

In multivariate time series prediction tasks, the inter- and intra-variable relations have significant influence on prediction outcomes. In many engineering and industrial scenarios, the multivariate time series also contain a large number of subjective influencing factors, such as settings and behaviors of users. Existing learning methods neglect the interactions of these subjective factors among variables. This leads to the learning of incorrect inter-variable influences, consequently yielding inaccurate prediction results. To address this challenge, we propose a Decoupled Casal Attention Network (DECA) for multivariate time series prediction from a spatiotemporal learning perspective. multivariate time series prediction. The causality decoupling module, based on the captured causal relations among variables, disentangles the subjective factors from the objective factors. Then the objective learning module utilizes an objective causal attention to capture objective cross-variable dependencies; while the subjective learning module utilizes a subjective causal graph attention to capture subjective influences. Finally, the prediction module fuses the multi-scale features of subjective and objective factors to produce predictions. The performance is evaluated using three benchmark datasets. Results indicate that, compared to state-of-the-art methods, DECA exhibits superior accuracy in multivariate time series prediction and can be effectively used for recommendations.

As a variant task of time-series segmentation, trajectory segmentation is a key task in the applications of transportation pattern recognition and traffic analysis. However, segmenting trajectory is faced with challenges of implicit patterns and sparse results. Although deep neural networks have tremendous advantages in terms of high-level feature learning performance, deploying as a blackbox seriously limits the real-world applications. Providing explainable segmentations has significance for result evaluation and decision making. Thus, in this article, we address trajectory segmentation by proposing a Bayesian Encoder-Decoder Network (BED-Net) to provide accurate detection with explainability and references for the following active-learning procedures. BED-Net consists of a segmentation module based on Monte Carlo dropout and an explanation module based on uncertainty learning that provides results evaluation and visualization. Experimental results on both benchmark and real-world datasets indicate that BED-Net outperforms the rival methods and offers excellent explainability in the applications of trajectory segmentation.

Given a location-based social network, how to find the communities that are highly relevant to query users and have top overall scores in multiple attributes according to user preferences? Typically, in the face of such a problem setting, we can model the network as a multi-attributed road-social network, in which each user is linked with location information and d (≥1) numerical attributes. In practice, user preferences (i.e., weights) are usually inherently uncertain and can only be estimated with bounded accuracy, because a human user is not able to designate exact values with absolute precision. Inspired by this, we introduce a normative community model suitable for multi-criteria decision making, called multi-attributed community (MAC), based on the concepts of k-core and a novel dominance relationship specific to preferences. Given uncertain user preferences, namely, an approximate representation of weights, the MAC search reports the exact communities for each of the possible weight settings. We devise an elegant index structure to maintain the dominance relationships, based on which two algorithms are developed to efficiently compute the top-j MACs. The efficiency and scalability of our algorithms and the effectiveness of MAC model are demonstrated by extensive experiments on both real-world and synthetic road-social networks.

Ventricular fibrillation is the most lethal arrhythmia. At present, the treatment of ventricular fibrillation is commonly received after the onset of the disease, which mainly depends on external defibrillation and drug-assisted therapy. Although activity of heartbeats can be described and analyzed using the most popular technique ECG (electrocardiogram), there is still no widely recognized prediction methods for ventricular fibrillation. Therefore, in this paper, in order to realize warning of ventricular fibrillation, we focus on the detection of atrial fibrillation and ventricular flutter, which are the arrhythmias often occurring before ventricular fibrillation. We propose a frequency-domain LSTM (Long Short-Term Memory), which uses heartbeat waves transformed from the original time domain into the frequency domain as input. Furthermore, to address the problem of big ECG data training efficiency and scalability, we also provide an implementation of our method under the distributed computing framework MapReduce in the Spark cluster. Experimental results indicate that our method achieves excellent classification performance compared with rival methods.