No.

Title Author

1 LSTM-ReGAT: A network-centric approach for Chao Zhong, Wei Du, Wei Xu,
cryptocurrency price trend prediction Qianhui Huang, Yinuo Zhao,
Mingming Wang,

2 Centralized decomposition approach in LSTM for Eunho Koo, Geonwoo Kim,

Bitcoin price prediction, Expert Systems with
Applications,

3 Enhancing Digital Cryptocurrency Trading Price Dawei Shang, Ziyu Guo, Hui Wang,
Prediction with an Attention-Based Convolutional
and Recurrent Neural Network Approach: The Case
of Ethereum,
4 Machine learning for cryptocurrency market Patrick Jaquart, Sven Köpke, Christof
prediction and trading, Weinhardt,

5 A survey of deep learning applications in Junhuan Zhang, Kewei Cai, Jiaqi Wen,
cryptocurrency,

6 A Deep Learning-based Cryptocurrency Price Mohil Maheshkumar Patel, Sudeep

Prediction Scheme for Financial Institutions, Tanwar, Rajesh Gupta, Neeraj Kumar,

7 Cryptocurrency price forecasting – A comparative Ahmed Bouteska, Mohammad Zoynul

analysis of ensemble learning and deep learning Abedin, Petr Hajek, Kunpeng Yuan,
methods,
 8 Attention-based CNN–LSTM for high-frequency Peng Peng, Yuehong Chen, Weiwei
multiple cryptocurrency trend prediction, Lin, James Z. Wang,

9 Cryptocurrency PricePrediction using LSTM S. Deo, G. Kumar, S. Soni and G. Singh

10 LSTM Based Sentiment Analysis for Huang, Xin and Zhang, Wenbin and
Cryptocurrency Prediction Tang, Xuejiao and Zhang, Mingli and
Surbiryala, Jayachander and Iosifidis,
Vasileios and Liu, Zhen and Zhang, Ji

11 Cryptocurrency Price Prediction Using LSTM and A. -A. Encean and D. Zinca
GRU Networks
Abstract

Predicting price trends of cryptocurrencies is a challenging task due to the highly speculative cryptocurrency market.
mainly investigate predictors such as historical trading data, macroeconomic development, and public interests in cryp
for price trend prediction while ignoring the predictive role of the relations between cryptocurrencies. In fact, the pric
of a cryptocurrency may be affected by those of other cryptocurrencies, thus, incorporating cryptocurrency interrelatio
further improve the prediction performance. In this paper, we propose a novel end-to-end model for price trend predic
long short-term memory (LSTM) and relationwise graph attention network (ReGAT), in which both individual crypto
features and cryptocurrency relations are considered. Specifically, a cryptocurrency network is built by using shared f
between cryptocurrencies. Further, LSTM is used to profile sequential patterns of individual cryptocurrency features.
extract network features, ReGAT is designed to aggregate information conveyed by different types of relations while
automatically differentiating their importance. Finally, the conventional and network features are concatenated to pred
trend. The effectiveness of LSTM-ReGAT is validated using real-world cryptocurrency market data. The trading simu
Bitcoin and portfolios reveal that our model obtains the highest profits. Our study provides insightful implications for
decision support in cryptocurrency market.

It has been reported that integrating time-series decomposition methods and neural network models improves financia
prediction performance. Despite its practical importance, the prediction performance of cryptocurrency prices, includ
at the tail domain of the label distribution is generally less successful than the mean performance across the entire dom
label distribution. In order to enhance the overall predictive performance of the Bitcoin price, we propose the Centrali
Distribution (CCD) as a novel input data filtering mechanism that significantly improves both the tail performance an
performance by mitigating the extreme bimodality inherent in Bitcoin price. The combination of CCD and the Weigh
Stretching (WES) loss function, which imposes different penalties depending on the label distribution, outcomes in an
performance gain. In the Long-Short Term Memory (LSTM) and the Singular Spectrum Analysis (SSA) decompositi
the CCD-WES strategy outperforms the native experiment by 11.5% and 22.5% Root Mean Square Error (RMSE) ga
whole and extreme domains of the label, respectively.

To predict Ethereum price fluctuations, this study proposes a new two-stage Machine Learning approach using an imp
convolutional neural network and a recurrent neural network framework, integrating an attention mechanism-based di
function algorithm. We construct a dataset and perform model training, fitting, and forecasting. The results indicate th
with traditional neural networks and time-series models such as GRU and ARIMA, respectively, this approach can ef
the data information of digital cryptocurrency and improve the prediction accuracy and interpretability of attention-ba
allocation functions. This study contributes to the literature by offering a new approach for stakeholders.
We employ and analyze various machine learning models for daily cryptocurrency market prediction and trading. We
models to predict binary relative daily market movements of the 100 largest cryptocurrencies. Our results show that a
models make statistically viable predictions, whereby the average accuracy values calculated on all cryptocurrencies r
52.9% to 54.1%. These accuracy values increase to a range from 57.5% to 59.5% when calculated on the subset of pr
with the 10% highest model confidences per class and day. We find that a long-short portfolio strategy based on the p
the employed LSTM and GRU ensemble models yields an annualized out-of-sample Sharpe ratio after transaction cos
and 3.12, respectively. In comparison, the buy-and-hold benchmark market portfolio strategy only yields a Sharpe rat
These results indicate a challenge to weak form cryptocurrency market efficiency, albeit the influence of certain limit
cannot be entirely ruled out

This study aims to comprehensively review a recently emerging multidisciplinary area related to the application of de
methods in cryptocurrency research. We first review popular deep learning models employed in multiple financial app
scenarios, including convolutional neural networks, recurrent neural networks, deep belief networks, and deep reinfor
learning. We also give an overview of cryptocurrencies by outlining the cryptocurrency history and discussing primar
representative currencies. Based on the reviewed deep learning methods and cryptocurrencies, we conduct a literature
deep learning methods in cryptocurrency research across various modeling tasks, including price prediction, portfolio
construction, bubble analysis, abnormal trading, trading regulations and initial coin offering in cryptocurrency. Moreo
discuss and evaluate the reviewed studies from perspectives of modeling approaches, empirical data, experiment resu
specific innovations. Finally, we conclude this literature review by informing future research directions and foci for d
in cryptocurrency

A cryptocurrency is a network-based digital exchange medium, where the records are secured using strong cryptogra
algorithms such as Secure Hash Algorithm 2 (SHA-2) and Message Digest 5 (MD5). It uses blockchain technology to
transactions secure, transparent, traceable, and immutable. Due to these properties, the cryptocurrencies have gained p
almost all the sectors especially in financial sectors. Though, cryptocurrencies are getting recognition form the approv
but still, the uncertainty and dynamism in their prices risk the investments substantially. Cryptocurrency price predict
become a trending research topic globally. Many machine learning and deep learning algorithms such as Gated Recur
(GRU), Neural Networks (NN), and Long short-term memory (LSTM) have been used by the researchers to predict a
the factors affecting the cryptocurrency prices. In this paper, a LSTM and GRU-based hybrid cryptocurrency predicti
proposed, which focuses on only two cryptocurrencies, namely Litecoin and Monero. The results depict that the propo
accurately predicts the prices with high accuracy, revealing that the scheme can be applicable in various cryptocurren
predictions.

Cryptocurrency price forecasting is attracting considerable interest due to its crucial decision support role in investme
Large fluctuations in non-stationary cryptocurrency prices motivate the urgent need for accurate forecasting models. T
seasonal effects and the need to meet a number of unrealistic requirements make it difficult to make accurate forecast
traditional statistical methods, leaving machine learning, particularly ensemble and deep learning, as the best technolo
area of cryptocurrency price forecasting. This is the first work to provide a comprehensive comparative analysis of en
learning and deep learning forecasting models, examining their relative performance on various cryptocurrencies (Bit
Ethereum, Ripple, and Litecoin) and exploring their potential trading applications. The results of this study reveal tha
recurrent unit, simple recurrent neural network, and LightGBM methods outperform other machine learning methods,
the naive buy-and-hold and random walk strategies. This can effectively guide investors in the cryptocurrency market
With the price of Bitcoin, Ethereum, and many other cryptocurrencies climbing, the cryptocurrency market has becom
popular investment area in recent years. Unlike other relatively more stable financial derivatives, the cryptocurrency m
high volatility which requires a high-frequency prediction model for quantitative trading. However, the excessive num
trading becomes a critical issue due to the instability of the prediction results and high error rate. To relieve such a pro
on the observation of high-frequency data, we use local minimum series to replace the original series and propose a m
triple trend labeling method that reduces the number of trades by potentially influencing the training of the model. In
new attention-based CNN–LSTM model for multiple cryptocurrencies (ACLMC) is proposed to optimize model effec
exploiting correlations across frequencies and currencies, and to smooth out the investment risk associated with predi
by supporting simultaneous multi-currency predictions. Experiments show that our labeling method with ACLMC ca
much better financial metrics and fewer number of transactions than traditional baselines.

This paper focuses on utilizing the Long Short-Term Memory (LSTM) algorithm to forecast the prices of cryptocurre
implemented through the model, which will allow us to keep tabs on the fluctuating prices of the various coins that ar
As various cryptocurrencies are available for trading and investment and they are regarded as a type of digital asset. B
trading and investing should be done after the knowledge of the market and trends, So, to solve this problem there sho
analysis before trading any pair. Our goal is to present accurate prediction based on the dataset of the selected cryptoc
Hence, the proposed model is built and it concentrate mostly on technical analysis of the crypto coins which is the ma
and most suitable for the prediction models. The proposed LSTM model achieves a variance regression score of 0.95
of the Bitcoin which serves as a representative cryptocurrency that is used in the model.

Recent studies in big data analytics and natural language processing develop automatic techniques in analyzing sentim
social media information. In addition, the growing user base of social media and the high volume of posts also provid
sentiment information to predict the price fluctuation of the cryptocurrency. This research is directed to predicting the
price movement of cryptocurrency by analyzing the sentiment in social media and finding the correlation between the
previous work has been developed to analyze sentiment in English social media posts, we propose a method to identif
sentiment of the Chinese social media posts from the most popular Chinese social media platform Sina-Weibo. We de
pipeline to capture Weibo posts, describe the creation of the crypto-specific sentiment dictionary, and propose a long
memory (LSTM) based recurrent neural network along with the historical cryptocurrency price movement to predict t
trend for future time frames. The conducted experiments demonstrate the proposed approach outperforms the state of
regressive based model by 18.5\% in precision and 15.4\% in recall

cryptocurrencies are increasingly used by investors as assets, mainly for short term. Therefore, in the financial sector,
important and can be done using Deep Learning techniques. This paper presents results on the prediction of three cryp
that see high volumes but are discussed in very few papers. In this paper we used LSTM (Long Short-Term Memory)
also GRU (Gated Recurrent Unit). These are applied to three important cryptocurrencies, namely DOGECoin, Ripple
OMG Network. The results presented prove that LSTM and GRU can be successful in predicting the daily closing va
cryptocurrencies.
Year DOI Plus Minus

2023 https://doi.org/10.1016/ -Considers both individual -Complexity of the model

j.dss.2023.113955 cryptocurrency features and may make it difficult to
cryptocurrency relations implement or understand
-Uses LSTM for sequential patterns -May require significant
and ReGAT for network features computational resources
-Validated using real-world data
-Obtains highest profits in trading
simulations

2024 https://doi.org/10.1016/j.eswa.2023.121401
- Improves both tail performance -- Focuses only on
and overall performance Bitcoin, may not
- Mitigates extreme bimodality in generalize to other
Bitcoin price cryptocurrencies
-Combines CCD and WES loss -Complexity of the
function for additional performance approach may limit
gain practical implementation
- Significant RMSE gains in whole
and extreme domains

2024 https://doi.org/10.1016/j.frl.2024.105846
- Two-stage Machine Learning - Focuses only on
approach Ethereum, may not apply
- Integrates attention mechanism- to other cryptocurrencies
based distribution function - May require extensive
algorithm data and computational
- Improves prediction accuracy and resources
interpretability
- Offers a new approach for
stakeholders
 - Analyzes various machine learning
2022 https://doi.org/10.1016/j.jfds.2022.12.001 - Challenges weak form
models market efficiency, which
- Statistically viable predictions for 100
largest cryptocurrencies may be controversial
- Higher accuracy for high-confidence - Influence of limits to
predictions arbitrage not fully ruled
- Long-short portfolio strategy yields out
high Sharpe ratios

2024 https://doi.org/10.1016/ - Comprehensive review of deep - As a survey, doesn't

j.isci.2023.108509 learning in cryptocurrency research propose new models or
- Covers multiple modeling tasks methods
and cryptocurrencies<br>- - May become outdated
Evaluates studies from various quickly in this fast-moving
perspectives field
- Provides future research directions

2020 https://doi.org/10.1016/j.jisa.2020.102583
- Focuses on Litecoin and Monero -Limited to only two
- Uses LSTM and GRU-based cryptocurrencies in the
hybrid prediction scheme study
-Claims high accuracy in price -May not account for all
predictions factors affecting
-Potentially applicable to various cryptocurrency prices
cryptocurrencies

2024 https://doi.org/10.1016/j.irfa.2023.103055
-- Comprehensive comparative -Does not propose new
analysis methods
- Examines multiple - Performance may vary in
cryptocurrencies different market
-Identifies top-performing methods conditions
(GRU, RNN, LightGBM
-Provides guidance for investors
2024 https://doi.org/10.1016/j.eswa.2023.121520
- Uses local minimum series for - May be complex to
more stable predictions implement
- Proposes triple trend labeling - High-frequency trading
method approach may not be
- Optimizes model effects across suitable for all investors |
frequencies and currencies
- Achieves better financial metrics
with fewer transactions

2023 https://doi.org/10.1109/ICCCNT56998.2023.10306885
- Focuses on technical analysis - Limited to Bitcoin as
- Achieves high variance regression representative
score for Bitcoin cryptocurrency
- Aims for accurate predictions - May not consider
based on dataset external market factors

2021 https://doi.org/10.1007/978-3-030-73200-4_47
- Incorporates social media - May be limited to
sentiment analysis Chinese market sentiment
- Focuses on Chinese social media - Effectiveness may vary
(Sina-Weibo) for different
- Outperforms auto-regressive cryptocurrencies
models
- Combines sentiment with
historical price data

2022 https://doi.org/10.1109/ISETC56213.2022.10010329
- Applies to less-studied - Limited to daily
cryptocurrencies (DOGE, XRP, predictions
OMG) - May not account for
- Uses both LSTM and GRU models sudden market shifts
- Successful in predicting daily
closing values