Time series forecasting has become a critical task in data engineering, with the volume of time series data projected to reach 180 ZB by 2025. While traditional forecasting models are typically constrained to single domains, missing opportunities for transferring temporal patterns across different domains. Through analysis, we observe that time series from different domains, despite their distinct statistical characteristics, can be fundamentally understood through temporal dependency patterns, which manifest as either long-term dependencies ( like trends and cycles) or short-term dependencies ( like fluctuations and abrupt changes). This observation motivates us to rethink cross-domain modeling from the dependency preferences perspective. We propose LSTPO, a novel framework that captures cross-domain commonalities through temporal dependency preferences and leverages a meta-learning-based approach to prevent cross-domain training forgetting. LSTPO dynamically models changes in preference over time and swiftly adapts to preference variations across different domains, enabling robust cross-domain forecasting. Through extensive experimental evaluations, we have shown that LSTPO substantially outperforms state-of-the-art forecasting methods while enhancing model transferability under few-shot learning conditions. The source code will be made publicly available upon acceptance.

Text-Attributed Heterogeneous Graphs (TAHGs) integrate topological relationships with rich textual node attributes, offering expressive representations for complex multi-faceted data. While recent methods jointly leverage textual and structural information, they still face two critical limitations: (i) existing approaches are constrained to neighborhood modeling, failing to capture semantic dependencies in higher-order topologies; (ii) current techniques exhibit inadequate unified alignment strategies, limiting dynamic interaction between cross modalities. To address these challenges, we propose SATH, a self-supervised information aggregation model for TAHGs, designed to effectively leverage textual and structural information within TAHGs. SATH aggregates higher-order neighbor textual attributes through comparative learning, and dynamically aligns these attributes to higher-order topologies through a unified strategy. This approach integrates both types of information effectively, enhancing the expressiveness and discriminative capability of the learned node representations in downstream tasks. Extensive experiments on real-world datasets demonstrate that SATH significantly outperforms baseline models while eliminating the need for manual meta-path design or text feature concatenation. It also improves efficiency and scalability on large-scale TAHGs, achieving superior representation quality in TAHG-based tasks.

Multivariate Time Series Forecasting (MTSF) has long been a key research focus. Traditionally, these studies assume a fixed number of variables, but in real-world applications, Cyber-Physical Systems often expand as new sensors are deployed, increasing variables in MTSF. In light of this, we introduce a novel task, Expanding-variate Time Series Forecasting (EVTSF). This task presents unique challenges, specifically (1) handling inconsistent data shapes caused by adding new variables, and (2) addressing imbalanced spatio-temporal learning, where expanding variables have limited observed data due to the necessity for timely operation. To address these challenges, we propose STEV, a flexible spatio-temporal forecasting framework. STEV includes a new Flat Scheme to tackle the inconsistent data shape issue, which extends the graph-based spatio-temporal modeling architecture into 1D space by flattening the 2D samples along the variable dimension, making the model variable-scale-agnostic while still preserving dynamic spatial correlations through a holistic graph. Additionally, we introduce a novel Spatio-temporal Focal Learning strategy that incorporates a negative filter to resolve potential conflicts between contrastive learning and graph representation, and a focal contrastive loss as its core to guide the framework to focus on optimizing the expanding variables. To evaluate the effectiveness of STEV, we benchmark EVTSF performance on three real-world datasets from various domains and compare it against three potential solutions employing state-of-the-art (SOTA) MTSF models tailored for EVSTF. Experimental results show that STEV significantly outperforms its competitors, especially in handling expanding variables. Notably, STEV, with only 5% of observations during the expanding period, is on par with SOTA MTSF models trained with complete data. Further exploration of various expanding scenarios underscores the generalizability of STEV in real-world applications.

Spatio-temporal forecasting of future values of spatially correlated time series is important across many cyber-physical systems (CPS). Recent studies offer evidence that the use of graph neural networks to capture latent correlations between time series holds a potential for enhanced forecasting. However, most existing methods rely on predefined or self-learning graphs, which are either static or unintentionally dynamic, and thus cannot model the time-varying correlations that exhibit trends and periodicities caused by the regularity of the underlying processes in CPS. To tackle such limitation, we propose Time-aware Graph Structure Learning (TagSL), which extracts time-aware correlations among time series by measuring the interaction of node and time representations in high-dimensional spaces. Notably, we introduce time discrepancy learning that utilizes contrastive learning with distance-based regularization terms to constrain learned spatial correlations to a trend sequence. Additionally, we propose a periodic discriminant function to enable the capture of periodic changes from the state of nodes. Next, we present a Graph Convolution-based Gated Recurrent Unit (GCGRU) that jointly captures spatial and temporal dependencies while learning time-aware and node-specific patterns. Finally, we introduce a unified framework named Time-aware Graph Convolutional Recurrent Network (TGCRN), combining TagSL, and GCGRU in an encoder-decoder architecture for multi-step spatiotemporal forecasting. We report on experiments with TGCRN and popular existing approaches on five real-world datasets, thus providing evidence that TGCRN is capable of advancing the state-of-the-art. We also cover a detailed ablation study and visualization analysis, offering detailed insight into the effectiveness of time-aware structure learning.

Federated Learning (FL) has emerged as a promising collaborative learning paradigm that enables to train machine learning models across decentralized devices, while keeping the training data localized to preserve user privacy. However, the heterogeneity in both decentralized training data and distributed computing resources has posed significant challenges to the design of effective and efficient FL schemes. Most existing solutions either focus on tackling a single type of heterogeneity, or are unable to fully support model heterogeneity with low communication overhead, fast convergence, and good interpretability. In this paper, we present CloREF, a novel rule-based collaborative learning framework that allows devices in FL to use completely different local learning models to cater to both data and resource heterogeneity. In CloREF, each rule is represented as a linear model, which provides good interpretability. Each participating device chooses a local model and trains it using its local data. The decision boundary of each trained local model is then approximated using a set of rules, which effectively bridges the gap arising from model heterogeneity. All participating devices collaborate to select the optimal set of rules as the global model, employing evolutionary optimization to effectively fuse the knowledge acquired from all local models. Experimental results on both synthesized and real-world datasets demonstrate that the rules generated by our proposed method can mimic the behaviors of various learning models with high fidelity ($\gt $0.95 in most tests), and CloREF gives competitive performance in accuracy, AUC, and communication overhead, compared with both the best-performing model trained centrally and several state-of-the-art model-heterogeneous federated learning schemes.

With the digitization of modern cities, large data volumes and powerful computational resources facilitate the rapid update of intelligent models deployed in smart cities. Continual learning (CL) is a novel machine learning paradigm that constantly updates models to adapt to changing environments, where the learning tasks, data, and distributions can vary over time. Our survey provides a comprehensive review of continual learning methods that are widely used in smart city development. The content consists of three parts: 1) Methodology-wise. We categorize a large number of basic CL methods and advanced CL frameworks in combination with other learning paradigms including graph learning, spatial-temporal learning, multi-modal learning, and federated learning. 2) Application-wise. We present numerous CL applications covering transportation, environment, public health, safety, networks, and associated datasets related to urban computing. 3) Challenges. We discuss current problems and challenges and envision several promising research directions. We believe this survey can help relevant researchers quickly familiarize themselves with the current state of continual learning research used in smart city development and direct them to future research trends.

Urban flow analysis is an essential research for smart city construction, in which urban flow pattern analysis focuses on the continuous state of urban flow. How to mine, store and reuse traffic patterns from urban multi-source heterogeneous big data is challenging. Therefore, this paper proposes a knowledge mining network for regional flow pattern to mine and store the urban flow pattern. The proposed model consists of two modules. In the first module, the features of the region and its flow pattern are extracted as the entity and relation, respectively. In the second module, POI features are modeled to enhance the embedding representation of relation and entity. Based on the translation distance method, the knowledge triplets of regional flow patterns are mined. Finally, the proposed model is compared with some benchmark methods using Chengdu Didi order and POI datasets. Experimental results show that the proposed model is effective. In addition, the knowledge triplets are visualized and some application examples are introduced.

The International Classification of Diseases (ICD) is a standard for categorizing physical conditions, which has been widely used for analyzing clinical data and monitoring health issues. Manual ICD coding takes a long time and is vulnerable to errors, so people pay more and more attention to the application of deep neural networks in ICD automatic coding. However, there is still no comprehensive review of these studies and prospects for further research. This paper is not limited to the study of deep neural networks, but gives a formal definition of ICD coding problems, and then systematically reviews the existing literature on how to design deep neural networks to address the four major challenges of ICD coding tasks. This paper also summarizes the public data sets and future research directions, to provide a guidance for the research of ICD coding in medical field.