Revisiting the Modifiable Areal Unit Problem in Deep Traffic Prediction with Visual Analytics

Zeng, Wei; Lin, Chengqiao; Lin, Juncong; Jiang, Jincheng; Xia, Jiazhi; Turkay, Çağatay; Chen, Wei

doi:10.1109/tvcg.2020.3030410

Cited by 29 publications

(14 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…With the rapid development of big data technology, combining the protection and research of traditional culture with big data technology is a big trend. There are many studies based on big data technology for traditional culture, and they have got good results [ 11 ]. Color is the most intuitive and prominent feature that can distinguish things' surface characteristics.…”

Section: Related Workmentioning

confidence: 99%

“…The color histogram method reflects the proportion of color information of different images in the image color space and is an effective way to research color information. At the same time, the color histogram is based on other coordinate systems and color spaces [ 11 ]. It can be used to study color and spatial information by calculating RGB, CIEL ∗a∗b , HSV, etc.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

A Neural Network Model for Color Element Data Analysis for Urban Spatial Environment

Xia

2022

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

In this paper, a CNN model for color element data analysis of the urban spatial environment is constructed through an in-depth study of color element data analysis. This paper investigates a high-order structure formed by a few nodes; it proposes a motif-based graph autoencoder MODEL, combining redefined first- and second-order similarities and perfectly integrating motif structure and autoencoder. The algorithm first proposes an efficient graph transformation method to add the influence of central nodes. It then offers a primary awareness mechanism to aggregate the information of noncentral neighbors. Cen GCN_D and Cen GCN_E outperform the latest algorithms in node classification, link prediction, node clustering, and network visualization. As the number of network layers increases, the advantages of these two variants become progressively more prominent. This paper uses a support vector machine to implement classification validation based on CNN. The experimental results show that when 450 images are randomly selected as training data, the classification accuracy obtained by using the features of different CNN output layers is distributed between 91.4% and 95.2%. When the training set of the experiment reaches more than 300, the accuracy can exceed 90%, and the experimental results corresponding to different training sets a more stable trend. Finally, the trained classifier model is obtained in this thesis, which achieves the purpose of fast classification prediction based on CNN for color element data analysis of urban spatial environments.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

A Neural Network Model for Color Element Data Analysis for Urban Spatial Environment

Xia

2022

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

show abstract

“…For example, ref. [71] showed how MAUP can lead to perturbations in the convolution-based residual neural networks used for urban traffic prediction. Thus, the investigation of such effects on spatial machine learning is critically needed.…”

Section: Scalementioning

confidence: 99%

Machine Learning of Spatial Data

Nikparvar

Thill

2021

IJGI

View full text Add to dashboard Cite

Properties of spatially explicit data are often ignored or inadequately handled in machine learning for spatial domains of application. At the same time, resources that would identify these properties and investigate their influence and methods to handle them in machine learning applications are lagging behind. In this survey of the literature, we seek to identify and discuss spatial properties of data that influence the performance of machine learning. We review some of the best practices in handling such properties in spatial domains and discuss their advantages and disadvantages. We recognize two broad strands in this literature. In the first, the properties of spatial data are developed in the spatial observation matrix without amending the substance of the learning algorithm; in the other, spatial data properties are handled in the learning algorithm itself. While the latter have been far less explored, we argue that they offer the most promising prospects for the future of spatial machine learning.

show abstract

“…Accuracy is of primary concern for short-term traffic flow prediction, i.e., output predictions shall be close to ground truths. In addition, this work also considers spatial autocorrelation as a key performance indicator, as spatial units of locally high nonstationarity are more likely to produce high prediction errors [42].…”

Section: Problem Definition (Short-term Traffic Flow Prediction)mentioning

confidence: 99%

Modeling Spatial Nonstationarity via Deformable Convolutions for Deep Traffic Flow Prediction

Zeng¹,

Lin²,

Liu³

et al. 2021

IEEE Trans. Knowl. Data Eng.

Self Cite

View full text Add to dashboard Cite

Deep neural networks are being increasingly used for short-term traffic flow prediction. Existing convolution-based approaches typically partition an underlying territory into grid-like spatial units, and employ standard convolutions to learn spatial dependence among the units. However, standard convolutions with fixed geometric structures cannot fully model the nonstationary characteristics of local traffic flows. To overcome the deficiency, we introduce deformable convolution that augments the spatial sampling locations with additional offsets, to enhance the modeling capability of spatial nonstationarity. On this basis, we design a deep deformable convolutional residual network, namely DeFlow-Net, that can effectively model global spatial dependence, local spatial nonstationarity, and temporal periodicity of traffic flows. Furthermore, to fit better with convolutions, we suggest to first aggregate traffic flows according to pre-conceived regions of interest, then dispose to sequentially organized raster images for network input. Extensive experiments on real-world traffic flows demonstrate that DeFlow-Net outperforms existing solutions using standard convolutions, and spatial partition by pre-conceived regions further enhances the performance. Finally, we demonstrate the advantage of DeFlow-Net in maintaining spatial autocorrelation, and reveal the impacts of partition shapes and scales on deep traffic flow prediction.

show abstract

Revisiting the Modifiable Areal Unit Problem in Deep Traffic Prediction with Visual Analytics

Cited by 29 publications

References 41 publications

A Neural Network Model for Color Element Data Analysis for Urban Spatial Environment

A Neural Network Model for Color Element Data Analysis for Urban Spatial Environment

Machine Learning of Spatial Data

Modeling Spatial Nonstationarity via Deformable Convolutions for Deep Traffic Flow Prediction

Contact Info

Product

Resources

About