Learning Optimal Q-Function Using Deep Boltzmann Machine for Reliable Trading of Cryptocurrency

Seok Jun Bu; Sung Bae Cho

doi:10.1007/978-3-030-03493-1_49

Learning Optimal Q-Function Using Deep Boltzmann Machine for Reliable Trading of Cryptocurrency

Seok Jun Bu, Sung Bae Cho

College of Computing

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

7 Citations (Scopus)

Abstract

The explosive price volatility from the end of 2017 to January 2018 shows that bitcoin is a high risk asset. The deep reinforcement algorithm is straightforward idea for directly outputs the market management actions to achieve higher profit instead of higher price-prediction accuracy. However, existing deep reinforcement learning algorithms including Q-learning are also limited to problems caused by enormous searching space. We propose a combination of double Q-network and unsupervised pre-training using Deep Boltzmann Machine (DBM) to generate and enhance the optimal Q-function in cryptocurrency trading. We obtained the profit of 2,686% in simulation, whereas the best conventional model had that of 2,087% for the same period of test. In addition, our model records 24% of profit while market price significantly drops by −64%.

Original language	English
Title of host publication	Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings
Editors	Hujun Yin, Paulo Novais, David Camacho, Antonio J. Tallón-Ballesteros
Publisher	Springer Verlag
Pages	468-480
Number of pages	13
ISBN (Print)	9783030034924
DOIs	https://doi.org/10.1007/978-3-030-03493-1_49
Publication status	Published - 2018
Event	19th International Conference on Intelligent Data Engineering and Automated Learning, IDEAL 2018 - Madrid, Spain Duration: 2018 Nov 21 → 2018 Nov 23

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11314 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	19th International Conference on Intelligent Data Engineering and Automated Learning, IDEAL 2018
Country/Territory	Spain
City	Madrid
Period	18/11/21 → 18/11/23

Bibliographical note

Funding Information:
This research was supported by Korea Electric Power Corporation. (Grant number: R18XA05).

Funding Information:
This research was supported by Korea Electric Power Corporation. (Grant

Publisher Copyright:
© 2018, Springer Nature Switzerland AG.

All Science Journal Classification (ASJC) codes

Theoretical Computer Science
Computer Science(all)

Access to Document

10.1007/978-3-030-03493-1_49

Cite this

Bu, S. J., & Cho, S. B. (2018). Learning Optimal Q-Function Using Deep Boltzmann Machine for Reliable Trading of Cryptocurrency. In H. Yin, P. Novais, D. Camacho, & A. J. Tallón-Ballesteros (Eds.), Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings (pp. 468-480). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11314 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-03493-1_49

Bu, Seok Jun ; Cho, Sung Bae. / Learning Optimal Q-Function Using Deep Boltzmann Machine for Reliable Trading of Cryptocurrency. Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings. editor / Hujun Yin ; Paulo Novais ; David Camacho ; Antonio J. Tallón-Ballesteros. Springer Verlag, 2018. pp. 468-480 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{8ea4c420a39d48d0be0d09ee421370f2,

title = "Learning Optimal Q-Function Using Deep Boltzmann Machine for Reliable Trading of Cryptocurrency",

abstract = "The explosive price volatility from the end of 2017 to January 2018 shows that bitcoin is a high risk asset. The deep reinforcement algorithm is straightforward idea for directly outputs the market management actions to achieve higher profit instead of higher price-prediction accuracy. However, existing deep reinforcement learning algorithms including Q-learning are also limited to problems caused by enormous searching space. We propose a combination of double Q-network and unsupervised pre-training using Deep Boltzmann Machine (DBM) to generate and enhance the optimal Q-function in cryptocurrency trading. We obtained the profit of 2,686% in simulation, whereas the best conventional model had that of 2,087% for the same period of test. In addition, our model records 24% of profit while market price significantly drops by −64%.",

author = "Bu, {Seok Jun} and Cho, {Sung Bae}",

note = "Funding Information: This research was supported by Korea Electric Power Corporation. (Grant number: R18XA05). Funding Information: This research was supported by Korea Electric Power Corporation. (Grant Publisher Copyright: {\textcopyright} 2018, Springer Nature Switzerland AG.; 19th International Conference on Intelligent Data Engineering and Automated Learning, IDEAL 2018 ; Conference date: 21-11-2018 Through 23-11-2018",

year = "2018",

doi = "10.1007/978-3-030-03493-1_49",

language = "English",

isbn = "9783030034924",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "468--480",

editor = "Hujun Yin and Paulo Novais and David Camacho and Tall{\'o}n-Ballesteros, {Antonio J.}",

booktitle = "Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings",

address = "Germany",

}

Bu, SJ & Cho, SB 2018, Learning Optimal Q-Function Using Deep Boltzmann Machine for Reliable Trading of Cryptocurrency. in H Yin, P Novais, D Camacho & AJ Tallón-Ballesteros (eds), Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11314 LNCS, Springer Verlag, pp. 468-480, 19th International Conference on Intelligent Data Engineering and Automated Learning, IDEAL 2018, Madrid, Spain, 18/11/21. https://doi.org/10.1007/978-3-030-03493-1_49

Learning Optimal Q-Function Using Deep Boltzmann Machine for Reliable Trading of Cryptocurrency. / Bu, Seok Jun; Cho, Sung Bae.
Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings. ed. / Hujun Yin; Paulo Novais; David Camacho; Antonio J. Tallón-Ballesteros. Springer Verlag, 2018. p. 468-480 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11314 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Learning Optimal Q-Function Using Deep Boltzmann Machine for Reliable Trading of Cryptocurrency

AU - Bu, Seok Jun

AU - Cho, Sung Bae

N1 - Funding Information: This research was supported by Korea Electric Power Corporation. (Grant number: R18XA05). Funding Information: This research was supported by Korea Electric Power Corporation. (Grant Publisher Copyright: © 2018, Springer Nature Switzerland AG.

PY - 2018

Y1 - 2018

N2 - The explosive price volatility from the end of 2017 to January 2018 shows that bitcoin is a high risk asset. The deep reinforcement algorithm is straightforward idea for directly outputs the market management actions to achieve higher profit instead of higher price-prediction accuracy. However, existing deep reinforcement learning algorithms including Q-learning are also limited to problems caused by enormous searching space. We propose a combination of double Q-network and unsupervised pre-training using Deep Boltzmann Machine (DBM) to generate and enhance the optimal Q-function in cryptocurrency trading. We obtained the profit of 2,686% in simulation, whereas the best conventional model had that of 2,087% for the same period of test. In addition, our model records 24% of profit while market price significantly drops by −64%.

AB - The explosive price volatility from the end of 2017 to January 2018 shows that bitcoin is a high risk asset. The deep reinforcement algorithm is straightforward idea for directly outputs the market management actions to achieve higher profit instead of higher price-prediction accuracy. However, existing deep reinforcement learning algorithms including Q-learning are also limited to problems caused by enormous searching space. We propose a combination of double Q-network and unsupervised pre-training using Deep Boltzmann Machine (DBM) to generate and enhance the optimal Q-function in cryptocurrency trading. We obtained the profit of 2,686% in simulation, whereas the best conventional model had that of 2,087% for the same period of test. In addition, our model records 24% of profit while market price significantly drops by −64%.

UR - http://www.scopus.com/inward/record.url?scp=85057076158&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85057076158&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-03493-1_49

DO - 10.1007/978-3-030-03493-1_49

M3 - Conference contribution

AN - SCOPUS:85057076158

SN - 9783030034924

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 468

EP - 480

BT - Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings

A2 - Yin, Hujun

A2 - Novais, Paulo

A2 - Camacho, David

A2 - Tallón-Ballesteros, Antonio J.

PB - Springer Verlag

T2 - 19th International Conference on Intelligent Data Engineering and Automated Learning, IDEAL 2018

Y2 - 21 November 2018 through 23 November 2018

ER -

Bu SJ, Cho SB. Learning Optimal Q-Function Using Deep Boltzmann Machine for Reliable Trading of Cryptocurrency. In Yin H, Novais P, Camacho D, Tallón-Ballesteros AJ, editors, Intelligent Data Engineering and Automated Learning – IDEAL 2018 - 19th International Conference, Proceedings. Springer Verlag. 2018. p. 468-480. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-03493-1_49

Learning Optimal Q-Function Using Deep Boltzmann Machine for Reliable Trading of Cryptocurrency

Abstract

Publication series

Other

Bibliographical note

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this