Spatio-temporal forecasting is crucial for sustainable urban development and societal decision-making. However, real-world spatio-temporal data often exhibit open-world missingness: missing rates and patterns evolve dynamically across time and space, severely disrupting dependencies and challenging accurate forecasting. Traditional methods universally overlook the dynamic nature of missingness, resulting in degraded predictive accuracy. To address this gap, we propose a novel Spatio-Temporal Missing-aware Mixture-of-Experts (STMMoE) architecture, equipped with a three-stage training strategy. STMMoE dynamically adapts to varying missing rates through a gating mechanism that selects specialized expert branches. The three-stage training strategy improves end-to-end forecasting performance by aligning the representations of complete and missing data. Extensive experiments on two real-world datasets show that our method achieves state-of-the-art performance.

To keep track of the current state of modern cities, various sensors are deployed throughout the cities to collect geo-sensory data. These data are used for intelligent applications, such as predicting pollution levels and issuing traffic warnings. However, there are a limited number of sensors of a specific type in the city, resulting in coarse-grained geo-sensory data. To obtain more detailed geo-sensory data and improve the quality of service for applications, it is necessary to infer the data at locations without sensors. This task is challenging due to two main factors: 1) only a few labeled locations with sensors, making them rare and sparse; 2) the spatio-temporal correlations are complex and dynamic.To overcome the above issues, we propose a novel method based on spatio-temporal contrastive learning for geo-sensory data inference (GSDI). The proposed method consists of two major parts, i.e., spatio-temporal data augmentations and a hybrid spatio-temporal representation network. We propose several spatio-temporal data augmentation methods to generate positive samples for learning robust representations via contrastive learning. We also design a hybrid spatio-temporal representation network to efficiently learn spatio-temporal representations for geo-sensory data inference. Furthermore, experiments on the real-world dataset show the advantages of our proposed methods. Our method achieves at most 5.26% improvement at MAE.

Traffic forecasting is crucial in the realm of intelligent urban planning, playing an essential role in route optimization, arrival time estimation, and the prevention of congestion-related incidents. Deep learning has catalyzed the development of advanced spatio-temporal graph neural networks (STGNNs) for traffic predictions. A key advantage of STGNNs is their capability to capture complex spatial correlations, thereby improving prediction accuracy. Nevertheless, the existing methodologies, including dynamical graph-based networks and attention-based architectures, tend to overlook regular traffic patterns, leading to overly complex neural networks with less interpretability and more computational costs, which could be a bottleneck for practical applications.To address the challenge, we propose a lightweight and user-friendly dynamic adaptive graph generator, termed the Vector-Quantized Graph Generator (VQGG), which can autonomously identify prevalent spatial patterns and construct the corresponding adaptive graphs efficiently for enhanced traffic forecasting. We have extensively tested the VQGG with two benchmark datasets by integrating it as a substitute for the adaptive graphs in baseline models with minor modifications. Extensive experimental results demonstrate that VQGG can reduce forecasting errors by approximately 2.59%, indicating a significant improvement in predictive performance.

Air pollution is a crucial issue affecting human health and livelihoods, as well as one of the barriers to economic growth. Forecasting air quality has become an increasingly important endeavor with significant social impacts, especially in emerging countries. In this paper, we present a novel Transformer termed AirFormer to predict nationwide air quality in China, with an unprecedented fine spatial granularity covering thousands of locations. AirFormer decouples the learning process into two stages: 1) a bottom-up deterministic stage that contains two new types of self-attention mechanisms to efficiently learn spatio-temporal representations; 2) a top-down stochastic stage with latent variables to capture the intrinsic uncertainty of air quality data. We evaluate AirFormer with 4-year data from 1,085 stations in Chinese Mainland. Compared to prior models, AirFormer reduces prediction errors by 5%∼8% on 72-hour future predictions. Our source code is available at https://github.com/yoshall/airformer.

The natural gas tightly correlates with our everyday life. However, driven by gray incomes, some users are prone to stealing gas by refitting the equipment without permission. Especially for the boiler room users in winter, this phenomenon appears more rampant. Traditional gas-theft detection methods highly rely on the on-site inspection, where exists ineffective and randomness. With the rapidly deployed IoT sensors, we can collect real-time gas consumption data to analyze users’ behavior patterns, where the gas-theft suspects could be discovered early and accurately. In this paper, we propose a data-driven approach, named SVOC, to detect gas-theft suspects among boiler room users. Our approach consists of a scenario-based data quality detection algorithm, a deformation-based normality detection algorithm, and an One-Class Support Vector Machine (OCSVM) based anomaly detection algorithm. Specifically, considering the temporal proximity between the gas consumption and the outdoor temperature, the normality detection algorithm adopts a similarity-based deformation correlation to detect normal boiler room users out of abnormal ones. Then, we employ OCSVM as the anomaly detection algorithm to capture various features across multiple data sources, aiming to distinguish gas-theft suspects from the remaining irregular users. Here, the detected normal and abnormal users are fed into the OCSVM for training and prediction, respectively, which can overcome the label scarcity problem. We conduct extensive experiments on a real-world dataset during one heating season. The results demonstrate distinct advantages of our approach over various baselines. We have developed a real-time system on the cloud, providing daily gas-theft suspects for gas companies.

Purchase prediction is an essential task in both online and offline retail industry, especially during major shopping festivals, when strong promotion boosts consumption dramatically. It is important for merchants to forecast such surge of sales and have better preparation. This is a challenging problem, as the purchase patterns during shopping festivals are significantly different from usual cases and also rare in historical data. Most existing methods fail at this problem due to the extremely scarce data samples as well as the inability to capture the complex macroscopic spatio-temporal dependencies in a city. To address this problem, we propose the Spatio-Temporal Meta-learning Prediction (STMP) model for purchase prediction during shopping festivals. STMP is a meta-learning based spatio-temporal multi-task deep generative model. It adopts a meta-learning framework with few-shot learning capability to capture both spatial and temporal data representations. A generative component then uses the extracted spatio-temporal representation and input data to infer the prediction results. Extensive experiments demonstrate the meta-learning generalization ability of STMP. STMP outperforms baselines in all cases, which shows the effectiveness of our model.

Spatio-temporal graphs are important structures to describe urban sensory data, e.g., traffic speed and air quality. Predicting over spatio-temporal graphs enables many essential applications in intelligent cities, such as traffic management and environment analysis. Recently, many deep learning models have been proposed for spatio-temporal graph prediction and achieved significant results. However, designing neural networks requires rich domain knowledge and expert efforts. To this end, we study automated neural architecture search for spatio-temporal graphs with the application to urban traffic prediction, which meets two challenges: 1) how to define search space for capturing complex spatio-temporal correlations; and 2) how to learn network weight parameters related to the corresponding attributed graph of a spatio-temporal graph.

Numerous sensors have been deployed in different geospatial locations to continuously and cooperatively monitor the surrounding environment, such as the air quality. These sensors generate multiple geo-sensory time series, with spatial correlations between their readings. Forecasting geo-sensory time series is of great importance yet very challenging as it is affected by many complex factors, i.e., dynamic spatio-temporal correlations and external factors. In this paper, we predict the readings of a geo-sensor over several future hours by using a multi-level attention-based recurrent neural network that considers multiple sensors' readings, meteorological data, and spatial data. More specifically, our model consists of two major parts: 1) a multi-level attention mechanism to model the dynamic spatio-temporal dependencies. 2) a general fusion module to incorporate the external factors from different domains. Experiments on two types of real-world datasets, viz., air quality data and water quality data, demonstrate that our method outperforms nine baseline methods.