LSTM-Based Dynamic Linguistic Decision-Making for Cryptocurrency Selection

Pablo Poblete-Arrué, Romina Torres, Víctor Salazar-Vasquez, Gustavo Gatica

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


Analyzing the overwhelming number of options available in the cryptocurrency market through technical analysis becomes unfeasible, whether expert investor or not. Similarly, challenging is the application of fundamental analysis in such markets. To address these complexities, this paper proposes a novel method to assist investors in navigating the overwhelming array of cryptocurrency options by employing a ‘buy-and-sell’ strategy. The approach incorporates a dynamic daily ranking system generated through LSTM neural network predictions and a dynamic linguistic decision-making model (DLDM). Simulated on a dataset of 68 cryptocurrencies observed from March to May 2018, the method surpasses state-of-the-art returns by over 300% when considering combinations of Day Profitability, Day Variability, and one of Open, Close, Low, or High attributes. Furthermore, by using a unitary constant as the third attribute, it achieves even higher returns, outperforming the state-of-the-art by more than 1700%. Comparatively, the proposed method easily outshines alternative strategies such as random selection, Bitcoin buy-and-hold, and equitable investment in all cryptocurrencies, which yielded returns of 3%, -34%, and 54%, respectively. The integration of LSTM predictions and DLDM showcases a potent tool for making informed decisions in the dynamic cryptocurrency market, especially crucial given the multitude of investment options and prevalence of non-expert investors. This paper presents a powerful approach to cryptocurrency investment, leveraging LSTM predictions and dynamic linguistic decision-making to provide investors with a competitive edge. The method's superior performance against published strategies showcases its potential for effectively tackling the complexities in cryptocurrency market, benefiting investors, despite experienced or not.

Original languageEnglish
Title of host publicationProceedings of International Conference on Information Technology and Applications - ICITA 2023
EditorsAbrar Ullah, Sajid Anwar, Davide Calandra, Raffaele Di Fuccio
PublisherSpringer Science and Business Media Deutschland GmbH
Number of pages14
ISBN (Print)9789819983230
Publication statusPublished - 2024
Event17th International Conference on Information Technology and Applications, ICITA 2023 - Lisbon, Portugal
Duration: 20 Oct 202222 Oct 2022

Publication series

NameLecture Notes in Networks and Systems
ISSN (Print)2367-3370
ISSN (Electronic)2367-3389


Conference17th International Conference on Information Technology and Applications, ICITA 2023


  • Cryptocurrencies investment
  • Decision-making
  • Linguistic decision model
  • Neural network

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Signal Processing
  • Computer Networks and Communications


Dive into the research topics of 'LSTM-Based Dynamic Linguistic Decision-Making for Cryptocurrency Selection'. Together they form a unique fingerprint.

Cite this