Convolution-Attention Model for Long-Term Traffic Forecasting 


Vol. 47,  No. 2, pp. 283-290, Feb.  2022
10.7840/kics.2022.47.2.283


PDF
  Abstract

In modern cities, traffic forecasting is one of the important problems, and various models have been studied to predict traffic flow. However, the traffic forecasting problem is a complex problem in that both temporal and spatial properties must be considered. Accordingly, various spatio-temporal neural network models which consider the two properties together are emerging recently. However, many previous studies have difficulty predicting long-term time series traffic flow. To improve predictive performance in the long-term, we build a model based on a dilated causal convolution and an attention model. We show that our model is effective for long-term prediction through experiments on real-world datasets. Experimental results show better performance than previous models such as dilated causal convolution-based models and attention models in long-term traffic forecasting.

  Statistics
Cumulative Counts from November, 2022
Multiple requests among the same browser session are counted as one view. If you mouse over a chart, the values of data points will be shown.


  Cite this article

[IEEE Style]

B. Choe, M. Lee, K. Jung, "Convolution-Attention Model for Long-Term Traffic Forecasting," The Journal of Korean Institute of Communications and Information Sciences, vol. 47, no. 2, pp. 283-290, 2022. DOI: 10.7840/kics.2022.47.2.283.

[ACM Style]

Byeongjin Choe, Minwoo Lee, and Kyomin Jung. 2022. Convolution-Attention Model for Long-Term Traffic Forecasting. The Journal of Korean Institute of Communications and Information Sciences, 47, 2, (2022), 283-290. DOI: 10.7840/kics.2022.47.2.283.

[KICS Style]

Byeongjin Choe, Minwoo Lee, Kyomin Jung, "Convolution-Attention Model for Long-Term Traffic Forecasting," The Journal of Korean Institute of Communications and Information Sciences, vol. 47, no. 2, pp. 283-290, 2. 2022. (https://doi.org/10.7840/kics.2022.47.2.283)

Search
(IN TITLE, AUTHOR, ABSTRACT,KEYWORDS)

Advanced Search ▼

POPULAR KEYWORDS
(TOP 10 KEYWORDS)

Recent Publications
(LAST 3 YEARS)






pISSN : 1226 - 4717
eISSN : 2287 - 3880