Search

Crowd flow prediction is a vital problem for an intelligent transportation system construction in a smart city. It plays a crucial role in traffic management and behavioral analysis, thus it has raised great attention from many researchers. However, predicting crowd flows timely and accurately is a challenging task that is affected by many complex factors such as the dependencies of adjacent regions or recent crowd flows. Existing models mainly focus on capturing such dependencies in spatial or temporal domains and fail to model relations between crowd flows of distant regions. We notice that each region has a relatively fixed daily flow and some regions (even very far away from each other) may share similar flow patterns which show strong correlations among them. In this article, we propose a novel model named Double-Encoder which follows a general encoder–decoder framework for multi-step citywide crowd flow prediction. The model consists of two encoder modules named ST-Encoder and FR-Encoder to model spatial-temporal dependencies and daily flow correlations, respectively. We conduct extensive experiments on two real-world datasets to evaluate the performance of the proposed model and show that our model consistently outperforms state-of-the-art methods.

Air pollution is an important environmental issue of increasing concern, which impacts human health. Accurate air quality prediction is crucial for avoiding people suffering from serious air pollution. Most of the prior works focus on capturing the temporal trend of air quality for each monitoring station. Recent deep learning based methods also model spatial dependencies among neighboring stations. However, we observe that besides geospatially adjacent stations, the stations which share similar functionalities or consistent temporal patterns could also have strong dependencies. In this paper, we propose an Attentive Temporal Graph Convolutional Network (ATGCN) to model diverse inter-station relationships for air quality prediction of citywide stations. Specifically, we first encode three types of relationships among stations including spatial adjacency, functional similarity, and temporal pattern similarity into graphs. Then we design parallel encoding modules, which respectively incorporate attentive graph convolution operations into the Gated Recurrent Units (GRUs) to iteratively aggregate features from related stations with different graphs. Furthermore, augmented with an attention-based fusion unit, decoding modules with a similar structure to the encoding modules are designed to generate multi-step predictions for all stations. The experiments on two real-world datasets demonstrate the superior performance of our model beyond state-of-the-art methods.

The full coverage of Wi-Fi signals and the popularization of intelligent card systems provide a large volume of data that contain human mobility patterns. Effectively utilizing such data to make user behavior predictions finds useful applications such as predictive behavior analysis, personalized recommendation, and location-aware services. Existing methods for user behavior predictions merely capture temporal dependencies within individual historical records. We argue that user behaviors are largely affected by friends in their social circles and such influences are dynamic due to users’ dynamic social behaviors. In this paper, we propose a model named SDSIM which consists of three independent and complementary modules to jointly model the influences of user dynamic social behaviors, user demographics similarities, and individual-level behavior patterns. We construct time-evolving graphs to indicate user dynamic social behaviors and design a novel component named DSBcell which captures not only the social influences but also the regularity and periodicity of user social behaviors. We also construct a graph based on user similarities in demographics and generate a representation for each user. Experiments on two real-world datasets for multiple user behavior-related prediction tasks prove the effectiveness of our proposed model compared with state-of-the-art methods.

Prediction tasks about students such as predicting students' academic performances have practical real-world significance at both the student level and the college level. With the rapid construction of smart campuses, colleges not only offer residence and academic programs but also record students' daily life. The digital footprints provide an opportunity to offer better solutions for prediction tasks. In this paper, we aim to propose a general deep neural network which can jointly model student heterogeneous daily behaviors generated from digital footprints and social influence to deal with prediction tasks. To this end, we design a variant of LSTM and a novel attention mechanism to model the daily behavior sequence. The proposed LSTM is able to consider context information (e.g., weather conditions) while modeling the daily behavior sequence. The proposed attention mechanism can dynamically learn the different importance degrees of different days for every student. Based on behavior information, we propose an unsupervised way to construct a social network to model social influence. Moreover, we design a residual network based decoder to model the complex interactions between the features and get the predicted values such as future academic performances. Qualitative and quantitative experiments on two real-world datasets collected from a college have demonstrated the effectiveness of our model.

Traffic flow forecasting is significant to traffic management and public safety. However, it is a challenging problem, because of complex spatial and temporal dependencies. Many existing approaches adopt Graph Convolution Networks (GCN) to model spatial dependencies and recurrent neural networks (RNN) to model temporal dependencies, simultaneously. However, the existing approaches mainly use adjacency matrix or distance matrix to represent the correlations between adjacent road segments, which fail to capture dynamic spatial dependencies. Besides, these approaches ignore the lag influence caused by propagation times of traffic flows and cannot model the global aggregation effect of traffic flows. In response to the limitations of the existing approaches, we model local aggregation and global aggregation of traffic flows. We propose a novel model, called the Local and Global Spatial Temporal Network (LGSTN), to forecast the traffic flows on a road segment basis (instead of regions). We first construct time-dependent flow transfer graphs to capture dynamic spatial correlations among the local traffic flows of the adjacent road segments. Next, we adopt spatial-based GCNs to model local traffic flow aggregation. Then, we propose a Lag-gated LSTM to model global traffic flow aggregation by considering free-flow reachable time matrix. Experiments on two real-world datasets have demonstrated our proposed LGSTN considerably outperforms state-of-the-art traffic forecast methods.

For the emerging mobility-on-demand services, it is of great significance to predict passenger demands based on historical mobility trips towards better vehicle distribution. Prior works have focused on predicting next-step passenger demands at selected locations or hotspots. However, we argue that multi-step citywide passenger demands encapsulate both time-varying demand trends and global statuses, and hence are more beneficial to avoiding demand-service mismatching and developing effective vehicle distribution/scheduling strategies. Furthermore, we find that adaptations of single-step methods are unable to achieve robust prediction with high accuracy for further steps. In this paper, we propose an end-to-end deep neural network model to the prediction task. We employ an encoder-decoder framework based on convolutional and ConvLSTM units to identify complex features that capture spatiotemporal influence and pickup-dropoff interactions on citywide passenger demands. We introduce a multi-level attention model (global attention and temporal attention) to emphasize the effects of latent citywide mobility regularities and capture relevant temporal dependencies. We evaluate our proposed method using real-world mobility trips (taxis and bikes) and the experimental results show that our method achieves higher prediction accuracy than the state-of-the-art approaches.