Air pollution is one of the major problems of the modern world. The popularization and powerful functions of smartphone applications enable people to participate in urban sensing to better know about the air problems surrounding them. Data sampling is one of the most important problems that affect the sensing performance. In this paper, we propose an Adaptive Sampling Scheme for Urban Air Quality (AS-air) through participatory sensing. Firstly, we propose to find the pattern rules of air quality according to the historical data contributed by participants based on Apriori algorithm. Based on it, we predict the on-line air quality and use it to accelerate the learning process to choose and adapt the sampling parameter based on Q-learning. The evaluation results show that AS-air provides an energy-efficient sampling strategy, which is adaptive toward the varied outside air environment with good sampling efficiency.

Urban crowd flow prediction is very challenging for public management and planning in smart city applications. Existing work mostly focuses on spatial and temporal dependence based flow prediction that are not well suited for predictions of instantaneous flow change usually due to social emergency incidents and accidents. In this paper, we propose an Online to Offline Interaction based Dilated Casual Convolutional Neural Network framework (O2O-DCNN) for crowd flow prediction. Both online attention behavior and offline crowd shift factors are considered, in case to capture the dependence between them and make more accurate predictions especially for instantaneous flow variations. The online and offline features are processed by dilated casual convolutions and then put into CBOW model based full connected network to make interactions. The performance evaluations are based on realistic User Detail Record (UDR) dataset of a southern city in China provided by China Unicom. O2O-DCNN is compared with the other related baselines in terms of MASE and MAE. The results show that our framework is with much better accuracy, especially for instantaneous flow variation scenario.

In recent years, information popularity prediction of online social media plays a vital role in crisis early warning and malicious content propagation identification within public opinion management application scenarios. Existing work lacks effective mechanisms for interactive topology feature extraction among multiple correlated cascades, which contributes the propagation scale prediction especially at the early propagation stage. In this paper, we propose a Persistence augmented Graph Convolution Network framework (PT-GCN) to make popularity prediction defined as retweet number of information propagation. Persistence as the topological data analysis approach is utilized to measure and find out the salient topology structure feature. Based on it, we propose to use multi-dimensional cascade graphs to model the correlated cascades and then execute PT-GCN to form the interactive propagation features from correlated nodes in propagation and correlated cascades with persistence & contents. The performance evaluations are based on three datasets from Weibo and Twitter platform. By comparisons with the other related work, PT-GCN achieves much better efficiency in terms of MSLE and precision, which performs well for early stage propagation.

Mobile edge computing works well with the rapid growth of human-centric IoT applications by providing timely and in-situ data caching and processing. Edge oriented urban hotspot prediction is to predict the high traffic regions in case to provide basis for further deployment and replenishment of edge servers. Most of the existing hotspot analysis work is not well suited for human-centric IoT application scenarios especially for large-scale urban city areas. Instead, in this paper we propose edge-oriented hotspot prediction scheme based on information Interactive Graph Attention Networks (I-GAN) for urban city within the human-centric IoT and MEC application scenarios. I-GAN makes reasoning on the user service needs by learning the interactive relations of multifaceted human-centric factors: including online human behavior and offline crowd flow through graph attention networks. On the other hand, I-GAN makes representation of urban-scale edge service capability by graph embeddings through inductive learning with distributed multi-layer graph convolutions. Based on it, I-GAN makes matching between them and then predicts the edge-oriented hotspots with time variance. The performance evaluations are based on realistic dataset of a southern city of China provided by China Unicom. I-GAN is compared with the other related hotspot analysis schemes. The results show that I-GAN is with much better prediction accuracy with multiple time dimensions.

Epidemic forecasting and contact network reconstruction are inherently coupled, yet most existing methods address them independently, limiting performance and interpretability. In this work, we propose EpiLoop, a unified closed-loop framework that enables bidirectional enhancement between case forecasting and latent contact network inference. EpiLoop integrates two synergistic components: DualView-GraphLSTM, a spatiotemporal graph neural network that models infection dynamics and mobility patterns, and VGAE-CR, a variational graph autoencoder that reconstructs time-varying subpopulation-level contact networks. Crucially, EpiLoop establishes a feedback mechanism: reconstructed networks are fed back to refine future forecasts, while latent infection states from forecasting guide network reconstruction—enabling mutual co-refinement through iterative optimization. We evaluate EpiLoop on real-world COVID-19 trajectory data from four Chinese provinces, using a comprehensive metapopulation network. Results show that EpiLoop achieves state-of-the-art performance, outperforming decoupled baselines in both forecasting accuracy (lower MAE and RMSE) and contact reconstruction (AUC up to 88.8%) under partial observability and behavioral changes. By unifying prediction and mechanism discovery, EpiLoop advances epidemic modeling toward transparent, self-improving systems and generalizes to behavior-modulated dynamics beyond COVID-19.