2F-TP: Learning Flexible Spatiotemporal Dependency for Flexible Traffic Prediction

ABSTRACTThe effectiveness of local spatial dependence in shaping the population density distribution is investigated. Individual location preferences are modelled by considering the status-related features of a given spatial unit and its neighbours as well as local random spatial dependence. The novelty is framing such a dependence through conditionally autoregressive (CAR) census random effects that are added to a spatially lagged explanatory variable X (SLX) setting. The results not only confirm that controlling for the spatial dimension is relevant but also indicate that local spatial dependence warrants consideration when determining the population distribution of recent decades. In this respect, the framework turns out to be useful for the analysis of microdata in which individual relationships (in a same spatial unit) enforce local spatial dependence.

As the fundamental part of other Intelligent Transportation Systems (ITS) applications, short-term traffic volume prediction plays an important role in various intelligent transportation tasks, such as traffic management, traffic signal control and route planning. Although Neural-network-based traffic prediction methods can produce good results, most of the models can’t be explained in an intuitive way. In this paper, we not only proposed a model that increase the short-term prediction accuracy of the traffic volume, but also improved the interpretability of the model by analyzing the internal attention score learnt by the model. we propose a spatiotemporal attention mechanism-based multistep traffic volume prediction model (SAMM). Inside the model, an LSTM-based Encoder-Decoder network with a hybrid attention mechanism is introduced, which consists of spatial attention and temporal attention. In the first level, the local and global spatial attention mechanisms considering the micro traffic evolution and macro pattern similarity, respectively, are applied to capture and amplify the features from the highly correlated entrance stations. In the second level, a temporal attention mechanism is employed to amplify the features from the time steps captured as contributing more to the future exit volume. Considering the time-dependent characteristics and the continuity of the recent evolutionary traffic volume trend, the timestamp features and historical exit volume series of target stations are included as the external inputs. An experiment is conducted using data from the highway toll collection system of Guangdong Province, China. By extracting and analyzing the weights of the spatial and temporal attention layers, the contributions of the intermediate parameters are revealed and explained with knowledge acquired by historical statistics. The results show that the proposed model outperforms the state-of-the-art model by 29.51% in terms of MSE, 13.93% in terms of MAE, and 5.69% in terms of MAPE. The effectiveness of the Encoder-Decoder framework and the attention mechanism are also verified.

With the increasing development of intelligent detection devices, a vast amount of traffic flow data can be collected from intelligent transportation systems. However, these data often encounter issues such as missing and abnormal values, which can adversely affect the accuracy of future tasks like traffic flow forecasting. To address this problem, this paper proposes the Attention-based Spatiotemporal Generative Adversarial Imputation Network (ASTGAIN) model, comprising a generator and a discriminator, to conduct traffic volume imputation. The generator incorporates an information fuse module, a spatial attention mechanism, a causal inference module and a temporal attention mechanism, enabling it to capture historical information and extract spatiotemporal relationships from the traffic flow data. The discriminator features a bidirectional gated recurrent unit, which explores the temporal correlation of the imputed data to distinguish between imputed and original values. Additionally, we have devised an imputation filling technique that fully leverages the imputed data to enhance the imputation performance. Comparison experiments with several traditional imputation models demonstrate the superior performance of the ASTGAIN model across diverse missing scenarios.

As one of the most widely used applications of the Internet of Things (IoT), intelligent transportation system (ITS) is of great significance for urban traffic planning, traffic control, and traffic guidance. However, widespread traffic congestion occurs with the increased number of vehicles. The traffic flow prediction is a good idea for traffic congestion. Therefore, many schemes have been proposed for accurate and real-time traffic flow prediction, but there still exist many issues, including low accuracy, weak adaptability and inferior real-time. Meanwhile, the complex spatial and temporal dependencies in traffic flow are still challenging. To address the above issues, we propose a novel spatial-temporal Chebyshev graph neural network model (ST-ChebNet) for traffic flow prediction to capture the spatial-temporal features, which can ensure accurate traffic flow prediction. Concretely, we first add a fully connected layer to fuse the features of traffic data into a new feature to generate a matrix, and then the long short-term memory (LSTM) model is adopted to learn traffic state changes for capturing the temporal dependencies. Then, we use the Chebyshev graph neural network (ChebNet) to learn the complex topological structures in the traffic network for capturing the spatial dependencies. Eventually, the spatial features and the temporal features are fused to guarantee the traffic flow prediction. The experiments show that ST-ChebNet can make accurate and real-time traffic flow prediction compared with other eight baseline methods on real-world traffic data sets PeMS.

Traffic forecasting is a crucial aspect of modern Intelligent Transportation Systems (ITS) and the Internet of Vehicles (IoV), playing a vital role in improving the safety and efficiency of daily transportation activities. Despite the valuable contributions of traditional machine learning (ML) models and advanced deep learning (DL) techniques, there persist challenges in capturing the intricate spatial and temporal dependencies inherent in traffic flow. In response to these challenges, we present GRAPHITE, an innovative framework that combines Graph Neural Networks (GNNs) and Generative Adversarial Networks (GANs) to leverage generative reasoning for efficient traffic management. Our model seamlessly integrates historical traffic volume data collected by road sensors with local spatial information encoded through knowledge graphs (KGs) associated with each sensor. These KGs offer a structured representation of relationships between traffic sensors and points of interest (POIs) in their neighborhood, thereby enhancing the comprehension of the urban context and leading to more accurate traffic predictions. Extensive experiments conducted on diverse datasets underscore the efficacy of GRAPHITE. Notably, we achieved a maximum decrease in RMSE of 31.05% compared to GAN-GRU and a maximum increase in R2 of 8.15% compared to GAN-RNN, positioning GRAPHITE as a standout solution among the current state-of-the-art approaches. Our code is available at: https://github.com/MODAL-UNINA/GRAPHITE.

Accurate traffic flow prediction plays a vital role in intelligent transportation systems, helping traffic management departments maintain stable traffic order, reduce traffic congestion, and improve road safety. Existing prediction methods focus on dynamic modeling of the spatiotemporal dependencies of traffic flow, capturing the periodicity and spatial heterogeneity in traffic data. However, they still suffer from a lack of focus on the important local information in long-term predictions, leading to overly smooth results that fail to effectively capture sudden changes in traffic patterns. To address these limitations, we propose the sLSTM-Attention-Based Multi-Head Dynamic Graph Convolutional Network (sAMDGCN) model. Specifically, we extend sLSTM and introduce temporal trend-aware multi-head attention to jointly capture the complex temporal dependencies. We propose a multi-head dynamic graph convolutional network to capture a wider range of dynamic spatial dependencies. To validate the effectiveness of sAMDGCN, we perform extensive experiments on four real-world traffic flow datasets. Experimental results show that our proposed sAMDGCN model outperforms the advanced baseline methods in long-term traffic flow prediction tasks, demonstrating its superior performance in capturing complex and dynamic traffic patterns.

As a cornerstone of intelligent transportation systems, traffic flow prediction has garnered extensive research attention. However, traffic flow data exhibits pronounced spatio-temporal dynamics, rendering accurate traffic flow prediction a challenging endeavor. Existing methodologies compartmentalize the capture of temporal and spatial dependencies through distinct modules, overlooking latent spatio-temporal heterogeneity within traffic data. Furthermore, they often solely consider spatio-temporal correlations between adjacent historical time steps, neglecting the extraction of temporal dependencies across different time intervals. To address these issues, we propose a novel Multi-scale Fusion Dynamic Graph Neural Network (MFDGNN). MFDGNN employs Temporal Convolutional Networks (TCNs) to generate time series on multiple scales, subsequently extracting spatio-temporal features from these sequences to construct mutually independent fusion graphs. By leveraging fusion graphs to extract features from time series on multiple scales, our model adeptly captures spatio-temporal dependencies of traffic nodes across various time dimensions, thereby achieving enhanced precision in traffic flow prediction. Experimental results conducted on four publicly available datasets demonstrate the superiority of our model compared to other baseline approaches.

Traffic flow prediction is an important part of intelligent transportation system (ITS). Accurate traffic flow prediction information can provide reliable reference for traffic management decision-makers, effectively reduce traffic congestion, reduce vehicle exhaust pollution, save energy and facilitate the travel of the people. Most of the existing traffic flow prediction models simplify the traffic flow change process into a linear and stable process and ignore the influence of weather conditions on traffic flow, which will lead to large prediction deviation. In order to improve the accuracy of traffic flow prediction, this paper proposes an ensemble learning method based on ET and AdaBoost for fusing traffic and meteorological data under the non-stationary environment to predict the changes of traffic flow. The experiment is based on real traffic and meteorological data, and the results show that the proposed ensemble method is superior to the baseline methods.

The intelligent transportation system (ITS) has been proven capable of effectively addressing traffic congestion issues. For vehicles to perform effectively and improve mobility under the intelligent driving environment, real-time prediction of traffic speed is undoubtedly essential. Considering the complex spatiotemporal dependency inherent in traffic data, conventional prediction models encounter many limitations. To improve the prediction performance and investigate the temporal features, this study focuses on emerging deep neural networks (DNNs) using the Caltrans Performance Measurement System (PeMS) data. This research also establishes an intelligent driving environment in the simulation and compares the traditional car-following model with deep learning methods in terms of multiple performance metrics. The results indicate that both supervised learning and unsupervised learning are superior to the simulation-based model on the freeway, and the two deep learning networks are almost identical to one another. Besides, the result reveals that all models have their latent features for different time dimensions under the low traffic loads, transition states, and heavy traffic loads. This is critical in the application of prediction technologies in ITS. The findings can assist transportation researchers and traffic engineers in both traffic operation and management, such as bottleneck identification, platooning control, and route planning.

Dynamic multi-scale spatial-temporal graph convolutional network for traffic flow prediction

In metropolitan areas, common traffic issues include traffic congestion, traffic accidents, air pollution, and energy consumption occur. To resolve this issues, Intelligent Transportation Systems (ITS) have been evolved by many researchers. One of the important sub-systems in the development of ITS is a Traffic Management System (TMS) which attempts to reduce a traffic congestion. In fact, TMS itself relies on the estimation of traffic flow, therefore providing such an accurate traffic flow prediction is needed. For this reason, we aim to provide an accurate traffic flow prediction to facilitate this system. In this works, a Supervised Deep Learning Based Traffic Flow Prediction (SDLTFP) was proposed which is a type of fully-connected deep neural network (FC-DNN). Timely prediction is also a major issue in guaranteeing reliable traffic flow prediction. However, training a deep network could be time-consuming, and overfitting is might be happening, especially when feeding small data into the deep architecture. The network is learned perfectly during the training, but in testing with the new data, it could fail to generalize the model. We adopt the Batch Normalization (BN) and Dropout techniques to help the network training. SGD and momentum are carried out to update the weight. We then take advantage of open data as historical traffic data which are then used to predict future traffic flow with the proposed method and model above. Experiments show that the Mean Absolute Percentage Error (MAPE) for our traffic flow prediction is within 5 % using sample data and between 15% to 20% using out of the sample data. Training a deep network faster with BN and Dropout reduces the overfitting.

Dynamic traffic correlations based spatio-temporal graph convolutional network for urban traffic prediction

In recent years, various types of Intelligent Transportation Systems(ITSs) have produced a large amount of structured traffic flow data. Using these structured data and deep learning models to predict urban short-term traffic flow is an important focus of current research. In order to solve the problem of low accuracy of Long-Short Term Memory(LSTM) model in urban traffic flow prediction, this paper proposes an LSTM model based on attention mechanism. The proposed model can learn the importance of each past value to the current value from the long sequence of traffic data at the past moment, which makes it possible to extract more valuable features. Constructed a dataset using the traffic data in the core section of Wuhan for experiments, and the performance of the improved model is compared with the original LSTM model. The results show that when the input data sequence increases from 16 to 64, the MAPE of our proposed model is reduced by 3.76%, while the MAPE of the LSTM model is reduced by 1.51%, which proves the effectiveness of our proposed method.

Accurate traffic flow prediction is crucial for improving transportation efficiency. To improve the accuracy of traffic flow prediction, we developed a traffic flow prediction framework—namely, traffic flow multicomponent network—that appropriately processes the noise, volatility, and nonlinearity in traffic flow data. This framework comprises three components: a factor selection component, traffic flow decomposition component, and traffic flow prediction component. The factor selection component considers the dynamic effects of weather‐related, environmental, and spatiotemporal factors on traffic flow; it then extracts and analyzes factors exhibiting strong correlations with traffic flow. The traffic flow decomposition component optimizes the parameters of variational mode decomposition on the basis of the envelope entropy by using the sparrow search algorithm; it then transforms traffic flow into multiple intrinsic mode functions to enable accurate traffic flow prediction. Finally, the traffic flow prediction component constructs dynamic feature matrices by using a bidirectional gated recurrent unit model to identify relationships within the data. Moreover, it uses an attention mechanism to assign different weights to different features on the basis of the importance of these features to traffic flow prediction, thereby enabling the efficient processing of a large volume of data. The performance of the proposed framework was examined in experiments conducted on large volumes of traffic flow data with different time granularities. The results indicated that the proposed framework achieved high prediction accuracy and stability for various time granularities, data samples, dataset sizes, and noise conditions. Moreover, it generally outperformed existing traffic flow prediction models under all experimental conditions.

Building data-driven intelligent transportation is a significant task for establishing data-centric smart cities, and exceptionally efficient and accurate traffic flow prediction (TFP) is a crucial technology in constructing intelligent transportation systems (ITSs). To address the computation and storage problems of processing traffic flow big data with the centralized model on a traditional mining platform, we propose a distributed long short-term memory weighted model combined with a time window and normal distribution based on a MapReduce parallel processing framework in this paper, named as WND-LSTM. More specifically, under the Hadoop distributed computing platform, a distributed modeling framework of forecasting traffic flow on MapReduce is developed to solve the existing issues of storage and calculation in handling large-scale traffic flow data with the stand-alone learning model. Moreover, a distributed WND-LSTM model is presented on the MapReduce-based distributed modeling framework to enhance the accuracy, efficiency, and scalability of short-term TFP. Finally, we forecast the traffic flow on the Sanlihe East Road of Beijing in China using the proposed WND-LSTM model with the real-world taxi trajectory big data. In particular, the extensively experimental results from a case study demonstrate that the MAPE value of WND-LSTM is 88.48%, 65.79%, 70.46%, 68.21%, 66.95%, 68.43%, and 70.41% lower than that of the autoregressive integrated moving average (ARIMA), logistical regression (LR), support vector regression (SVR), k-nearest neighbor (KNN), stacked autoencoders (SAEs), gated recurrent unit (GRU), and long short-term memory (LSTM), respectively, and achieves 71.25% accuracy improvement on average.