In the transformative landscape of smart cities, the integration of the cutting-edge web technologies into time series forecasting presents a pivotal opportunity to enhance urban planning, sustainability, and economic growth. The advancement of deep neural networks has significantly improved forecasting performance. However, a notable challenge lies in the ability of these models to generalize well to out-of-distribution (OOD) time series data. The inherent spatial heterogeneity and domain shifts across urban environments create hurdles that prevent models from adapting and performing effectively in new urban environments. To tackle this problem, we propose a solution to derive invariant representations for more robust predictions under different urban environments instead of relying on spurious correlation across urban environments for better generalizability. Through extensive experiments on both synthetic and real-world data, we demonstrate that our proposed method outperforms traditional time series forecasting models when tackling domain shifts in changing urban environments. The effectiveness and robustness of our method can be extended to diverse fields including climate modeling, urban planning, and smart city resource management.

In the data-driven era, recommendations have become indispensable across various systems. Graphs, as versatile data structures, shine at abstracting complex systems. Many real-world scenarios effortlessly translate into graphs, representing individuals and their relationships as nodes and edges. Link prediction, a cornerstone of recommendations, excels in forecasting future network connections based on current structures. Its applications span diverse domains, including social networks, biological networks, and network security. Previous studies have leveraged classification algorithms like logistic regression and random forest, often complemented by node embedding techniques, yielding impressive results in addressing the challenge of link prediction. Today's dynamic networks continually reshape connections, introducing new links and nodes while removing others. Furthermore, the inclusion of location information associated with nodes provides a new opportunity. Adapting models to this dynamism necessitates capturing spatial and temporal dependencies for sustained effectiveness. In this paper, we undertake a comprehensive evaluation of various algorithms for link prediction. Subsequently, we further enriched the continuous-time dynamic graph networks by incorporating essential location information. This strategic enhancement results in a remarkable performance improvement, highlighting the crucial role of location-based temporal data in improving recommendations. It emphasizes the untapped potential of location and temporal information in refining user recommendations within interconnected networks.

Granger causality, commonly used for inferring causal structures from time series data, has been adopted in widespread applications across various fields due to its intuitive explainability and high compatibility with emerging deep neural network prediction models. To alleviate challenges in better deciphering causal structures unambiguously from time series, the use of interventional data has become a practical approach. However, existing methods have yet to be explored in the context of imperfect interventions with unknown targets, which are more common and often more beneficial in a wide range of real-world applications. Additionally, the identifiability issues of Granger causality with unknown interventional targets in complex network models remain unsolved. Our work presents a theoretically-grounded method that infers Granger causal structure and identifies unknown targets by leveraging heterogeneous interventional time series data. We further illustrate that learning Granger causal structure and recovering interventional targets can mutually promote each other. Comparative experiments demonstrate that our method outperforms several robust baseline methods in learning Granger causal structure from interventional time series data.

Mining spatiotemporal mobility patterns is crucial for optimizing urban planning, enhancing transportation systems, and improving public safety by providing useful insights into human movement and behavior over space and time. As an unsupervised learning technique, time series clustering has gained considerable attention due to its efficiency. However, the existing literature has often overlooked the inherent characteristics of mobility data, including high-dimensionality, noise, outliers, and time distortions. This oversight can lead to potentially large computational costs and inaccurate patterns. To address these challenges, this paper proposes a novel neural network-based method integrating temporal autoencoder and dynamic time warping-based K-means clustering algorithm to mutually promote each other for mining spatiotemporal mobility patterns. Comparative results showed that our proposed method outperformed several time series clustering techniques in accurately identifying mobility patterns on both synthetic and real-world data, which provides a reliable foundation for data-driven decision-making. Furthermore, we applied the method to monthly county-level mobility data during the COVID-19 pandemic in the U.S., revealing significant differences in mobility changes between rural and urban areas, as well as the impact of public response and health considerations on mobility patterns.

Granger causality is widely used for causal structure discovery in complex systems from multivariate time series data. Traditional Granger causality tests based on linear models often fail to detect even mild non-linear causal relationships. Therefore, numerous recent studies have investigated non-linear Granger causality methods, achieving improved performance. However, these methods often rely on two key assumptions: causal sufficiency and known interventional targets. Causal sufficiency assumes the absence of latent confounders, yet their presence can introduce spurious correlations. Moreover, real-world time series data usually come from heterogeneous environments, without prior knowledge of interventions. Therefore, in practice, it is difficult to distinguish intervened environments from non-intervened ones, and even harder to identify which variables or timesteps are affected. To address these challenges, we propose Invariant Granger Causality (InvarGC), which leverages cross-environment heterogeneity to mitigate the effects of latent confounding and to distinguish intervened from non-intervened environments with edge-level granularity, thereby recovering invariant causal relations. In addition, we establish the identifiability under these conditions. Extensive experiments on both synthetic and real-world datasets demonstrate the competitive performance of our approach compared to state-of-the-art methods.