Urban traffic passenger flows prediction is practically important to facilitate many real applications including transportation management and public safety. Recently, deep learning based approaches are proposed learn the spatio-temporal characteristics of flows. However, it still very challenging handle some complex factors such as hybrid lines, mixed traffic, transfer stations, extreme weathers. Considering multi-channel irregularity properties urban in different a more efficient...

Dynamic Graph Neural Networks (DGNNs) have become one of the most promising methods for traffic speed forecasting. However, when adapting DGNNs forecasting, existing approaches are usually built on a static adjacency matrix (no matter predefined or self-learned) to learn spatial relationships among different road segments, even if impact two segments can be changeable dynamically during day. Moreover, future cannot only related with current speed, but also affected by other factors such as...

Accurate metro ridership prediction can guide passengers in efficiently selecting their departure time and transferring from station to station. An increasing number of deep learning algorithms are being utilized forecast due the development computational intelligence. However, limited efforts have been exerted consider spatiotemporal features, which important forecasting through methods, large-scale networks. To fill this gap, paper proposes a parallel architecture comprising convolutional...

Graph Convolutional Network (GCN) has been widely applied in transportation demand prediction due to its excellent ability capture non-Euclidean spatial dependence among station-level or regional demands. However, most of the existing research, graph convolution was implemented on a heuristically generated adjacency matrix, which could neither reflect real relationships stations accurately, nor multi-level demands adaptively. To cope with above problems, this paper provides novel...

Taxi and sharing bike bring great convenience to urban transportation. A lot of efforts have been made improve the efficiency taxi service or system by predicting next-period pick-up drop-off demand. Different from existing research, this paper is motivated following two facts: 1) From a micro view, an observed spatial demand at any time slot could be decomposed as combination many hidden bases; 2) macro multiple transportation demands are strongly correlated with each other, both spatially...

As a crucial component in intelligent transportation systems, traffic flow prediction has recently attracted widespread research interest the field of artificial intelligence (AI) with increasing availability massive mobility data. Its key challenge lies how to integrate diverse factors (such as temporal rules and spatial dependencies) infer evolution trend flow. To address this problem, we propose unified neural network called Attentive Traffic Flow Machine (ATFM), which can effectively...

In spatial crowdsourcing, requesters submit their task-related locations and increase the demand of a local area. The platform prices these tasks assigns workers to serve if are accepted by requesters. There exist mature pricing strategies which specialize in tackling imbalance between supply market. However, global optimization, should consider mobility workers; that is, any single worker can be potential for several areas, while it only true one area when assigned platform. hardness lies...

Representation learning on heterogeneous graphs aims to obtain meaningful node representations facilitate various downstream tasks, such as classification and link prediction. Existing graph methods are primarily developed by following the propagation mechanism of representations. There few efforts studying role relations for improving more fine-grained Indeed, it is important collaboratively learn semantic discern with respect different relation types. To this end, in paper, we propose a...

Spatiotemporal data mining plays an important role in air quality monitoring, crowd flow modeling, and climate forecasting. However, the originally collected spatiotemporal real-world scenarios is usually incomplete due to sensor failures or transmission loss. imputation aims fill missing values according observed underlying dependence of them. The previous dominant models impute autoregressively suffer from problem error accumulation. As emerging powerful generative models, diffusion...

In this paper, we present a novel approach to machine reading comprehension for the MS-MARCO dataset. Unlike SQuAD dataset that aims answer question with exact text spans in passage, defines task as answering from multiple passages and words are not necessary passages. We therefore develop an extraction-then-synthesis framework synthesize answers extraction results. Specifically, model is first employed predict most important sub-spans passage evidence, synthesis takes evidence additional...

Hashing has been proved as an attractive solution to approximate nearest neighbor search, owing its theoretical guarantee and computational efficiency. Though most of prior hashing algorithms can achieve low memory computation consumption by pursuing compact hash codes, however, they are still far beyond the capability learning discriminative functions from data with complex inherent structure among them. To address this issue, in paper, we propose a sensitive based on cluster prototypes,...

Precise traffic demand prediction could help government and enterprises make better management operation decisions by providing them with data-driven insights. However, it is a nontrivial effort to design an effective method due the spatial temporal characteristics of distributions, dynamics human mobility, impacts multiple environmental factors. To handle these problems, Dynamic Transition Convolutional Neural Network (DTCNN) proposed for purpose precise prediction. Particularly, transition...

Traffic flow prediction is crucial for urban traffic management and public safety. Its key challenges lie in how to adaptively integrate the various factors that affect changes. In this paper, we propose a unified neural network module address problem, called Attentive Crowd Flow Machine~(ACFM), which able infer evolution of crowd by learning dynamic representations temporally-varying data with an attention mechanism. Specifically, ACFM composed two progressive ConvLSTM units connected...

Accurate speed predictions for urban roads are highly important traffic monitoring and route planning, also help relieve the pressure of congestion. Many existing studies on prediction based convolutional neural networks, these have primarily focused capturing spatial proximity among different road segments. However, real cause spread congestion is connectivity segments, rather than their proximity. This makes it very challenging to improve accuracy. Using graph networks (GNNs), segments can...

Small sample sizes and imbalanced datasets have been two difficulties in previous traffic incident detection-related studies. Moreover, real-time characteristics of detection models must be improved to satisfy the needs management. In this study, a hybrid model is proposed address above problems. model, generative adversarial network (GAN) used expand size balance datasets, temporal spatially stacked autoencoder (TSSAE) extract spatial correlations flow detect incidents. Using real-world...