Traffic Prediction Based on Random Connectivity in Deep Learning with Long Short-Term Memory

Yuxiu Hua, Zhifeng Zhao, Zhiming Liu, Xianfu Chen, Rongpeng Li, Honggang Zhang

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

Abstract

Traffic prediction plays an important role in evaluating the performance of telecommunication networks and attracts intense research interests. A significant number of algorithms and models have been put forward to analyse traffic data and make prediction. In the recent big data era, deep learning has been exploited to mine the profound information hidden in the data. In particular, Long Short-Term Memory (LSTM), one kind of Recurrent Neural Network (RNN) schemes, has attracted a lot of attentions due to its capability of processing the long-range dependency embedded in the sequential traffic data. However, LSTM has considerable computational cost, which can not be tolerated in tasks with stringent latency requirement. In this paper, we propose a deep learning model based on LSTM, called Random Connectivity LSTM (RCLSTM). Compared to the conventional LSTM, RCLSTM makes a notable breakthrough in the formation of neural network, which is that the neurons are connected in a stochastic manner rather than full connected. We apply the RCLSTM to predict traffic and validate that the RCLSTM with even 35% neural connectivity still shows a satisfactory performance. When we gradually add training samples, the performance of RCLSTM becomes increasingly closer to the baseline LSTM. Moreover, for the input traffic sequences of enough length, the RCLSTM exhibits even superior prediction accuracy than the baseline LSTM.

Original languageEnglish
Title of host publication2018 IEEE 88th Vehicular Technology Conference (VTC-Fall)
PublisherInstitute of Electrical and Electronic Engineers IEEE
Number of pages6
ISBN (Electronic)978-1-5386-6358-5, 978-1-5386-6357-8
ISBN (Print)978-1-5386-6359-2
DOIs
Publication statusPublished - 2018
MoE publication typeA4 Article in a conference publication
Event88th IEEE Vehicular Technology Conference, VTC-Fall 2018 - Chicago, United States
Duration: 27 Aug 201830 Aug 2018
Conference number: 88

Publication series

NameIEEE Vehicular Technology Conference
Volume2018-August
ISSN (Print)1090-3038
ISSN (Electronic)2577-2465

Conference

Conference88th IEEE Vehicular Technology Conference, VTC-Fall 2018
Abbreviated titleVTC-Fall 2018
CountryUnited States
CityChicago
Period27/08/1830/08/18

Fingerprint

Memory Term
Connectivity
Traffic
Prediction
Baseline
Telecommunication Network
Recurrent neural networks
Training Samples
Recurrent Neural Networks
Learning
Long short-term memory
Deep learning
Neurons
Telecommunication networks
Latency
Computational Cost
Neuron
Neural Networks
Model-based
Neural networks

Keywords

  • big data
  • deep learning
  • LSTM
  • random connectivity
  • RNN
  • Traffic prediction

Cite this

Hua, Y., Zhao, Z., Liu, Z., Chen, X., Li, R., & Zhang, H. (2018). Traffic Prediction Based on Random Connectivity in Deep Learning with Long Short-Term Memory. In 2018 IEEE 88th Vehicular Technology Conference (VTC-Fall) [8690851] Institute of Electrical and Electronic Engineers IEEE. IEEE Vehicular Technology Conference Papers, Vol.. 2018-August https://doi.org/10.1109/VTCFall.2018.8690851
Hua, Yuxiu ; Zhao, Zhifeng ; Liu, Zhiming ; Chen, Xianfu ; Li, Rongpeng ; Zhang, Honggang. / Traffic Prediction Based on Random Connectivity in Deep Learning with Long Short-Term Memory. 2018 IEEE 88th Vehicular Technology Conference (VTC-Fall). Institute of Electrical and Electronic Engineers IEEE, 2018. (IEEE Vehicular Technology Conference Papers, Vol. 2018-August).
@inproceedings{3af090ed71b146da8f2e2fcc9bb4a921,
title = "Traffic Prediction Based on Random Connectivity in Deep Learning with Long Short-Term Memory",
abstract = "Traffic prediction plays an important role in evaluating the performance of telecommunication networks and attracts intense research interests. A significant number of algorithms and models have been put forward to analyse traffic data and make prediction. In the recent big data era, deep learning has been exploited to mine the profound information hidden in the data. In particular, Long Short-Term Memory (LSTM), one kind of Recurrent Neural Network (RNN) schemes, has attracted a lot of attentions due to its capability of processing the long-range dependency embedded in the sequential traffic data. However, LSTM has considerable computational cost, which can not be tolerated in tasks with stringent latency requirement. In this paper, we propose a deep learning model based on LSTM, called Random Connectivity LSTM (RCLSTM). Compared to the conventional LSTM, RCLSTM makes a notable breakthrough in the formation of neural network, which is that the neurons are connected in a stochastic manner rather than full connected. We apply the RCLSTM to predict traffic and validate that the RCLSTM with even 35{\%} neural connectivity still shows a satisfactory performance. When we gradually add training samples, the performance of RCLSTM becomes increasingly closer to the baseline LSTM. Moreover, for the input traffic sequences of enough length, the RCLSTM exhibits even superior prediction accuracy than the baseline LSTM.",
keywords = "big data, deep learning, LSTM, random connectivity, RNN, Traffic prediction",
author = "Yuxiu Hua and Zhifeng Zhao and Zhiming Liu and Xianfu Chen and Rongpeng Li and Honggang Zhang",
year = "2018",
doi = "10.1109/VTCFall.2018.8690851",
language = "English",
isbn = "978-1-5386-6359-2",
series = "IEEE Vehicular Technology Conference",
publisher = "Institute of Electrical and Electronic Engineers IEEE",
booktitle = "2018 IEEE 88th Vehicular Technology Conference (VTC-Fall)",
address = "United States",

}

Hua, Y, Zhao, Z, Liu, Z, Chen, X, Li, R & Zhang, H 2018, Traffic Prediction Based on Random Connectivity in Deep Learning with Long Short-Term Memory. in 2018 IEEE 88th Vehicular Technology Conference (VTC-Fall)., 8690851, Institute of Electrical and Electronic Engineers IEEE, IEEE Vehicular Technology Conference Papers, vol. 2018-August, 88th IEEE Vehicular Technology Conference, VTC-Fall 2018, Chicago, United States, 27/08/18. https://doi.org/10.1109/VTCFall.2018.8690851

Traffic Prediction Based on Random Connectivity in Deep Learning with Long Short-Term Memory. / Hua, Yuxiu; Zhao, Zhifeng; Liu, Zhiming; Chen, Xianfu; Li, Rongpeng; Zhang, Honggang.

2018 IEEE 88th Vehicular Technology Conference (VTC-Fall). Institute of Electrical and Electronic Engineers IEEE, 2018. 8690851 (IEEE Vehicular Technology Conference Papers, Vol. 2018-August).

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

TY - GEN

T1 - Traffic Prediction Based on Random Connectivity in Deep Learning with Long Short-Term Memory

AU - Hua, Yuxiu

AU - Zhao, Zhifeng

AU - Liu, Zhiming

AU - Chen, Xianfu

AU - Li, Rongpeng

AU - Zhang, Honggang

PY - 2018

Y1 - 2018

N2 - Traffic prediction plays an important role in evaluating the performance of telecommunication networks and attracts intense research interests. A significant number of algorithms and models have been put forward to analyse traffic data and make prediction. In the recent big data era, deep learning has been exploited to mine the profound information hidden in the data. In particular, Long Short-Term Memory (LSTM), one kind of Recurrent Neural Network (RNN) schemes, has attracted a lot of attentions due to its capability of processing the long-range dependency embedded in the sequential traffic data. However, LSTM has considerable computational cost, which can not be tolerated in tasks with stringent latency requirement. In this paper, we propose a deep learning model based on LSTM, called Random Connectivity LSTM (RCLSTM). Compared to the conventional LSTM, RCLSTM makes a notable breakthrough in the formation of neural network, which is that the neurons are connected in a stochastic manner rather than full connected. We apply the RCLSTM to predict traffic and validate that the RCLSTM with even 35% neural connectivity still shows a satisfactory performance. When we gradually add training samples, the performance of RCLSTM becomes increasingly closer to the baseline LSTM. Moreover, for the input traffic sequences of enough length, the RCLSTM exhibits even superior prediction accuracy than the baseline LSTM.

AB - Traffic prediction plays an important role in evaluating the performance of telecommunication networks and attracts intense research interests. A significant number of algorithms and models have been put forward to analyse traffic data and make prediction. In the recent big data era, deep learning has been exploited to mine the profound information hidden in the data. In particular, Long Short-Term Memory (LSTM), one kind of Recurrent Neural Network (RNN) schemes, has attracted a lot of attentions due to its capability of processing the long-range dependency embedded in the sequential traffic data. However, LSTM has considerable computational cost, which can not be tolerated in tasks with stringent latency requirement. In this paper, we propose a deep learning model based on LSTM, called Random Connectivity LSTM (RCLSTM). Compared to the conventional LSTM, RCLSTM makes a notable breakthrough in the formation of neural network, which is that the neurons are connected in a stochastic manner rather than full connected. We apply the RCLSTM to predict traffic and validate that the RCLSTM with even 35% neural connectivity still shows a satisfactory performance. When we gradually add training samples, the performance of RCLSTM becomes increasingly closer to the baseline LSTM. Moreover, for the input traffic sequences of enough length, the RCLSTM exhibits even superior prediction accuracy than the baseline LSTM.

KW - big data

KW - deep learning

KW - LSTM

KW - random connectivity

KW - RNN

KW - Traffic prediction

UR - http://www.scopus.com/inward/record.url?scp=85064888832&partnerID=8YFLogxK

U2 - 10.1109/VTCFall.2018.8690851

DO - 10.1109/VTCFall.2018.8690851

M3 - Conference article in proceedings

SN - 978-1-5386-6359-2

T3 - IEEE Vehicular Technology Conference

BT - 2018 IEEE 88th Vehicular Technology Conference (VTC-Fall)

PB - Institute of Electrical and Electronic Engineers IEEE

ER -

Hua Y, Zhao Z, Liu Z, Chen X, Li R, Zhang H. Traffic Prediction Based on Random Connectivity in Deep Learning with Long Short-Term Memory. In 2018 IEEE 88th Vehicular Technology Conference (VTC-Fall). Institute of Electrical and Electronic Engineers IEEE. 2018. 8690851. (IEEE Vehicular Technology Conference Papers, Vol. 2018-August). https://doi.org/10.1109/VTCFall.2018.8690851