This study aims to comprehensively review a recently emerging multidisciplinary area related to the application of deep learning methods in cryptocurrency research. We first review popular deep learning models employed in multiple financial application scenarios, including convolutional neural networks, recurrent neural networks, deep belief networks, and deep reinforcement learning. We also give an overview of cryptocurrencies by outlining the cryptocurrency history and discussing primary representative currencies. Based on the reviewed deep learning methods and cryptocurrencies, we conduct a literature review on 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. Moreover, we discuss and evaluate the reviewed studies from perspectives of modeling approaches, empirical data, experiment results and specific innovations. Finally, we conclude this literature review by informing future research directions and foci for deep learning in cryptocurrency.
Decentralized Finance, mushrooming in permissionless blockchains, has attracted a recent surge in popularity. Due to the transparency of permissionless blockchains, opportunistic traders can compete to earn revenue by extracting Miner Extractable Value (MEV), which undermines both the consensus security and efficiency of blockchain systems. The Flashbots bundle mechanism further aggravates the MEV competition because it empowers opportunistic traders with the capability of designing more sophisticated MEV extraction. In this paper, we conduct the first systematic study on DeFi MEV activities in Flashbots bundle by developing ActLifter, a novel automated tool for accurately identifying DeFi actions in transactions of each bundle, and ActCluster, a new approach that leverages iterative clustering to facilitate us to discover known/unknown DeFi MEV activities. Extensive experimental results show that ActLifter can achieve nearly 100% precision and recall in DeFi action identification, significantly outperforming state-of-the-art techniques. Moreover, with the help of ActCluster, we obtain many new observations and discover 17 new kinds of DeFi MEV activities, which occur in 53.12% of bundles but have not been reported in existing studies.
Sherin M. Omran, Wessam H. El-Behaidy, Aliaa A. A. Youssif
A cryptocurrency is a non-centralized form of money that facilitates financial transactions using cryptographic processes. It can be thought of as a virtual currency or a payment mechanism for sending and receiving money online. Cryptocurrencies have gained wide market acceptance and rapid development during the past few years. Due to the volatile nature of the crypto-market, cryptocurrency trading involves a high level of risk. In this paper, a new normalized decomposition-based, multi-objective particle swarm optimization (N-MOPSO/D) algorithm is presented for cryptocurrency algorithmic trading. The aim of this algorithm is to help traders find the best Litecoin trading strategies that improve their outcomes. The proposed algorithm is used to manage the trade-offs among three objectives: the return on investment, the Sortino ratio, and the number of trades. A hybrid weight assignment mechanism has also been proposed. It was compared against the trading rules with their standard parameters, MOPSO/D, using normalized weighted Tchebycheff scalarization, and MOEA/D. The proposed algorithm could outperform the counterpart algorithms for benchmark and real-world problems. Results showed that the proposed algorithm is very promising and stable under different market conditions. It could maintain the best returns and risk during both training and testing with a moderate number of trades.
Ferdiansyah Ferdiansyah, Raja Zahilah, Sit Hajar, Deris Stiawan
LSTM is a promising tool for predicting the stock exchange. Still, when the LSTM Model faces an anomaly problem with a dataset of Bitcoin that has hit more change in value by fluctuation, it can be a problem for producing good evaluation results such as RMSE. This research is an improvement over the discoveries of previous research. We tried another perspective besides using five years of historical data prices to predict a six-day value. We found that the results of RMSE were not very good but exhibited good results on MAPE as a comparison evaluation method. We are using the last six days to predict the next day. Logically, this dataset has good dataset stability, but the dataset has quite a significant minute-by-minute change in day-by-day value. Furthermore, CNN-LSTM was selected in this research to give another perspective and improve the results. The results were quite good and greatly improved previous research.
Kokulo K. Lawuobahsumo, Bernardina Algieri, Arturo Leccadito
Abstract This study aims to jointly predict conditional quantiles and tail expectations for the returns of the most popular cryptocurrencies (Bitcoin, Ethereum, Ripple, Dogecoin and Litecoin) using financial and macroeconomic indicators as explanatory variables. We adopt a Monotone Composite Quantile Regression Neural Network (MCQRNN) model to make one- and five-steps-ahead predictions of Value-at-Risk (VaR) and Expected Shortfall (ES) based on a rolling window and compare the performance of our model against the Historical simulation and the standard ARMA(1,1)-GARCH(1,1) model used as benchmarks. The superior set of models is then chosen by backtesting VaR and ES using a Model Confidence Set procedure. Our results show that the MCQRNN performs better than both benchmark models for jointly predicting VaR and ES when considering daily data. Models with the implied volatility index, treasury yield spread and inflation expectations sharpen the extreme return predictions. The results are consistent for the two risk measures at the 1% and 5% level both, in the case of a long and short position and for all cryptocurrencies.
Since cryptocurrencies are becoming more widely used and accepted in the financial system, precise price forecasting is essential for optimizing bitcoin investments. In this research study, we evaluated various machine learning models, including linear regression (LR), decision tree regression (DT), random forest regression (RF), support vector regression (SVR), gradient boosting regression (GB), adaboost regression, extreme gradient boosting regression (XGR), light gradientboosting regression (LGBM), k-nearest neighbors regression (KNN), ridge, andlasso. Additionally, we incorporated two deep learning (DL) models, namely artificial neural networks (ANN) and convolutional neural networks (CNN), to forecast daily bitcoin prices (BP). The initial data was obtained from Kaggle, a well-known platform for data science projects, and we applied the min-max scaler technique for consistent scaling during preprocessing. To assess the predictive capabilities of the models, we utilized regression metrics such as root mean square error (RMSE), mean absolute error (MAE), and correlation coefficient (R). Based on our findings, the CNN model demonstrated the highest effectiveness in predicting BPs among the DL models, with an RMSE of 0.0543, MAE of 0.0324, and an R value of 0.960. In the case of machine learning models, the RF model outperformed others, achieving an RMSE of 0.0246 and MAE of 0.0561. Investors, scholars, and decision-makers may all gain from these findings’ insightful revelations about BP forecasting. Developing these models further, investigating different preprocessing methods, and expanding the analysis to other cryptocurrencies might be the main goals of future research.
Bitcoin is always interesting to keep predicting where the next price movement will go. Bitcoin is the first and most influential Cryptocurrency on cryptocurrency price movements. Bitcoin is traded in many markets, the largest in Indonesia is Indodax. Indodax provides a document sharing API so that third parties can build applications that are able to process data on bitcoin price movements in real-time and continuously. This research shows how patterned datasets can be applied to monitor bitcoin price movements from the indodax market and show their effects on other cryptocurrency assets besides bitcoin. This research shows how data that is patterned and then processed using the minimum and maximum functions can provide 2 important information, namely the potential position of bitcoin when it is at the maximum and minimum points. The results of the patterned dataset formula are then compared to the movements of the 2 cryptocurrencies with the largest capitalization, namely BTC (Bitcoin) and ETH (Ethereum). The results of the comparison show that the use of patterned dataset formulas successfully shows important points in cryptocurrency trading, with simpler instructions.
In the realm of cryptocurrency forecasting, accurately predicting short-term Bitcoin log returns remains a challenging endeavor due to its inherent volatility and sensitivity to multifarious external factors. This study addresses this challenge by proposing an integrated approach that combines the capabilities of the TimesNet deep learning model with sentiment analysis techniques. TimesNet, specifically designed for time series data, has demonstrated proficiency in extracting salient patterns. When synergized with sentiment analysis, a more nuanced understanding of price determinants emerges. Preliminary results from our experiments indicate a significant enhancement in predictive accuracy within the Bitcoin market. Such advancements not only furnish investors and researchers with refined forecasting tools but also accentuate the burgeoning role of deep learning methodologies in the domain of financial forecasting.
El mercado de criptomonedas está experimentando un rápido crecimiento, lo que lo convierte en una alternativa potencialmente más lucrativa que los mercados financieros convencionales. No obstante, esta expansión va de la mano con una significativa volatilidad, presentando así un desafío crucial. En el contexto de esta tesis de maestría, se desarrollaron modelos de predicción de series temporales para el precio de cierre de Bitcoin mediante el uso de algoritmos de aprendizaje profundo, tales como LSTM y GRU. Además, se llevó a cabo una comparación con modelos tradicionales como ARIMA, con el propósito de analizar y evaluar su rendimiento.
Jinghua Wang, Geoffrey Ngene, Yan Shi, Ann Nduati Mungai
Policymakers and portfolio managers pay keen attention to sources of uncertainties that drive asset returns and volatility. The influence of uncertainty on Bitcoin has the potential to drive fluctuations in the entire cryptocurrency market. We investigate the predictability of thirteen economic policy uncertainty indices on Bitcoin returns. Using the Random Forest machine learning algorithm, we find that Singapore’s economic policy uncertainty (EPU) has the strongest predictive power on Bitcoin returns, followed by financial crisis (FC) uncertainty and world trade uncertainty (WTU). We further categorize these uncertainties into different groups. Interestingly, the predictability of uncertainty indices on Bitcoin returns within the international trade group is stronger compared to other uncertainty categories. Additionally, we observed that internet-based uncertainty measures have more predictive power of Bitcoin returns than newspaper- and report-based measures. These results are robust using various additional machine learning methods. We believe that these findings could be valuable for policymakers and portfolio managers when making decisions related to uncertainty drivers of cryptocurrency prices and returns.
Abstract Cryptocurrencies and Bitcoin, in particular, are prone to wild swings resulting in frequent jumps in prices, making them historically popular for traders to speculate. It is claimed in recent literature that Bitcoin price is influenced by sentiment about the Bitcoin system. Transaction, as well as the popularity, have shown positive evidence as potential drivers of Bitcoin price. This study introduces a bivariate jump-diffusion model to capture the dynamics of Bitcoin prices and the Bitcoin sentiment indicator, integrating trading volumes or Google search trends with Bitcoin price movements. We derive a closed-form solution for the Bitcoin price and the associated Black–Scholes equation for Bitcoin option valuation. The resulting partial differential equation for Bitcoin options is solved using an artificial neural network, and the model is validated with data from highly volatile stocks. We further test the model’s robustness across a broad spectrum of parameters, comparing the results to those obtained through Monte Carlo simulations. Our findings demonstrate the model’s practical significance in accurately predicting Bitcoin price movements and option values, providing a reliable tool for traders, analysts, and risk managers in the cryptocurrency market.
Rasoul Amirzadeh, Dhananjay Thiruvady, Asef Nazari, Mong Shan Ee
Despite advances in artificial intelligence-enhanced trading methods, developing a profitable automated trading system remains challenging in the rapidly evolving cryptocurrency market. This research focuses on developing a reinforcement learning (RL) framework to tackle the complexities of trading five prominent altcoins: Binance Coin, Ethereum, Litecoin, Ripple, and Tether. To this end, we present the CausalReinforceNet~(CRN) framework, which integrates both Bayesian and dynamic Bayesian network techniques to empower the RL agent in trade decision-making. We develop two agents using the framework based on distinct RL algorithms to analyse performance compared to the Buy-and-Hold benchmark strategy and a baseline RL model. The results indicate that our framework surpasses both models in profitability, highlighting CRN's consistent superiority, although the level of effectiveness varies across different cryptocurrencies.
We provide a comprehensive investigation into the profitability of technical trading methods applied to the cryptocurrency pairs BTC/USDT and ETH/USDT. By employing rigorous evaluations and incremental examinations, we address the pervasive issue of data-snooping bias that often plagues the evaluation of trading strategies. Our empirical results indicate the lack of profitable technical trading strategies in both the analysis sample and prediction sample periods, even after rigorous adjustments for data snooping. These findings highlight the difficulties associated with selecting profitable technical trading strategies in the dynamic and volatile cryptocurrency market. Market participants, including individual traders, institutional investors, and regulatory bodies, should take note of our findings when making investment decisions based on technical analysis.
Automated Market Makers (AMMs) are major centers of matching liquidity supply and demand in Decentralized Finance. Their functioning relies primarily on the presence of liquidity providers (LPs) incentivized to invest their assets into a liquidity pool. However, the prices at which a pooled asset is traded is often more stale than the prices on centralized and more liquid exchanges. This leads to the LPs suffering losses to arbitrage. This problem is addressed by adapting market prices to trader behavior, captured via the classical market microstructure model of Glosten and Milgrom. In this paper, we propose the first optimal Bayesian and the first model-free data-driven algorithm to optimally track the external price of the asset. The notion of optimality that we use enforces a zero-profit condition on the prices of the market maker, hence the name ZeroSwap. This ensures that the market maker balances losses to informed traders with profits from noise traders. The key property of our approach is the ability to estimate the external market price without the need for price oracles or loss oracles. Our theoretical guarantees on the performance of both these algorithms, ensuring the stability and convergence of their price recommendations, are of independent interest in the theory of reinforcement learning. We empirically demonstrate the robustness of our algorithms to changing market conditions.
Omar M. Ahmed, Lailan M. Haji, Ayah M. Ahmed, Nashwan M. Salih
The field of finance makes extensive use of real-time prediction of stock price tools, which are instruments that are put to use in the process of creating predictions. In this article, we attempt to predict the price of Bitcoin in a manner that is both accurate and reliable. Deep learning models, as opposed to more traditional methods, are used to manage enormous volumes of data and to generate predictions. The purpose of this research is to develop a method for predicting stock prices using the Hybrid Convolutional Recurrent Model (HCRM) architecture. This model architecture integrates the advantages of two separate deep learning models: The 1-Dimensional-Convolusional Neural Network (1D-CNN) and the Long-Short Term Memory (LSTM). The 1D-CNN is responsible for the feature extraction, while the LSTM is in charge of the temporal regression. The developed 1D-CNN-LSTM model has an outstanding performance in predicting stock values.
Market traders often earn income by buying and selling assets through market transactions with the goal of maximizing total returns. Since each asset purchased has a certain amount of gain and loss, traders need to develop a trading strategy, which may affect their total profit. In the case of gold and bitcoin, this paper develops an ARIMA daily price forecasting model and a trade strategy model based on price increases to help market traders develop optimal trading strategies for gold and bitcoin.
Predicting the prices of cryptocurrency owing to its volatility, instability, and other factors has been challenging; investors and traders especially in Nigeria have been on a constant look for a more reliable way of knowing market trends and prices and while there has been so many research conducted using deep learning, the results for which has fall short of what investors could called a strong predictor. This research reviewed the results of many works that had been done and proposed two types of recurrent neural network (RNNs) namely Long Short-Term Memory and Gated Recurrent Unit (GRU) for predicting the future prices of two of the most common crypto assets namely Bitcoin (BTC) and Ethereum (ETH), these two were selected based on their popularity, the volume traded and their market capitalization. The experiment was conducted in a GPU Jupyter Notebook environment and the performance of our experiment was evaluated on a test set using root mean square error (RMSE) and based on its values, the LSTM presented a better performance with rsme scores of 654.66, and 80.30 respectively for BTC and ETH as compared to GRU. The paper proceeds further to compare the results of this experiment with other related works and we discovered that its performance is a great improvement.
Quoc Minh Nguyen, Dat Tran, Juho Kanniainen, Alexandros Iosifidis · 5 authors
Many cryptocurrency brokers nowadays offer a va-riety of derivative assets that allow traders to perform hedging or speculation. This paper proposes an effective algorithm based on neural networks to take advantage of these investment products. The proposed algorithm constructs a portfolio that contains a pair of negatively correlated assets. A deep neural network, which outputs the allocation weight of each asset at a time interval, is trained to maximize the Sharpe ratio. A novel loss term is proposed to regulate the network's bias towards a specific asset, thus enforcing the network to learn an allocation strategy that is close to a minimum variance strategy. Extensive experiments were conducted using data collected from Binance spanning 19 months to evaluate the effectiveness of our approach. The backtest results show that the proposed algorithm can produce neural networks that are able to make profits in different market situations.
Cryptocurrency works similar to standard currency, however, virtual payments are made for goods and services without the intervention of any central authority. Many investors believe in and use Twitter tweets to guide their daily cryptocurrency trading. In this project, we investigated the feasibility of sentiment analysis and emotion for cryptocurrencies. For the study, we targeted (BTC) Bitcoin and collected related data. The data collection, cleaning were essential components of the study. Analysis of the sentiments about cryptocurrency is highly desirable to provide a holistic view of peoples' perceptions. In this regard, this study performs both sentiment analysis and emotion detection using the tweets related to the Bitcoin which are widely used for predicting the market prices of cryptocurrency. For increasing the efficacy of the analysis, a deep learning ensemble model LSTM-GRU is proposed that combines two recurrent neural networks applications. Comparatively, a larger number of people feel happy with the use of cryptocurrency, followed by fear and surprise emotions. The model achieves the highest performance for sentiment analysis with a 0.91 accuracy score and the highest emotion 0.83. Similarly, LSTM-GRU outperforms all other models in terms of correct and wrong predictions for both sentiment analysis 0.99 and emotion detection 0.98. Key Words: bitcoin, sentiment analysis, machine learning, cryptocurrencies, tweets
Cryptocurrency is a virtual payment instrument currently popular as an investment alternative. One type of cryptocurrency widely used as an investment is Bitcoin due to its high-profit potential and risk due to unstable exchange rate fluctuations. This high exchange rate fluctuation makes trading transactions in the crypto market speculative and highly volatile. To overcome this volatility factor, this research used the Generalized Autoregressive Conditional Heteroscedasticity forecasting method to describe the heteroscedasticity factor, as well as a Recurrent Neural Network (RNN) with long-short-term memory that has feedback in modeling sequential data for time series analysis. The two methods are combined to overcome the dependency of time series data in the long term and the heteroscedastic effect of the volatility of price changes. The results of the GARCH-LSTM hybrid model in this study show a Mean Absolute Percentage Error (MAPE) value of 15.69%. The accuracy value is obtained from the division of training data by 80% and testing data by 20%, with the number of neurons as many as three and epochs of 100 using the Adam optimizer. The MAPE accuracy results show a good prediction in predicting the value.
Purpose- This study aims to assess the effectiveness of technical analysis indicators used by investors in the cryptocurrency market for making informed decisions. Emphasizing the importance of accurate decision-making methods in financial markets, this research particularly focuses on the cryptocurrency market, which has gained significant attention among investors in recent years. Methodology- The study specifically examines technical analysis, a widely employed method in various financial markets, with a focus on its predictive capabilities concerning Bitcoin price forecasts. Leveraging advanced technologies, such as big data analysis and machine learning, the research utilizes daily trading data from January 1, 2017, to June 30, 2022, presenting technical indicators and their associated error margins. Findings- The study highlights the significance of using Weighted Moving Average (WMA) and Stochastic Oscillator (STO) indicators in combination, demonstrating that multiple indicators outperform individual ones. This research underscores the effectiveness of technical analysis methods in the cryptocurrency market, aiding the development of enhanced investment strategies. Conclusion- In conclusion, this study delves into the potency of technical analysis techniques employed by investors in cryptocurrency markets. The insights indicate that combining indicators and technical analysis methods holds promise for future investment strategies. It is essential to note that even the best method can lead to losses, as evidenced by the presence of error margins, and absolute profitability cannot be guaranteed through technical analysis methods. Keywords: Cryptocurrency, technical analysis, machine-learning, classification algorithms, investment. JEL Codes: C38, C55, G17
Muhammad Haziq Abdul Hadi, Nor Azuana Ramli, Qamar UI Islam
Predicting future prices of cryptocurrencies, including Bitcoin and Ethereum, presents a formidable challenge owing to their inherent volatility. This study applies Long Short-Term Memory (LSTM), a well-established recurrent neural network for time series forecasting, to predict Bitcoin and Ethereum values. Historical price data for both cryptocurrencies, sourced from Yahoo Finance, serves as the basis for analysis. The dataset undergoes an 80% training and 20% testing partition. Subsequently, LSTM models are developed and trained on both datasets. In parallel, the gated recurrent unit (GRU), recognized as an advanced variant of the LSTM model, is explored for comparative purposes. Performance evaluation utilizes fundamental metrics, including root mean squared error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). The results reveal an intriguing trend: both models exhibit superior performance when applied to the Ethereum dataset compared to the Bitcoin dataset. This observation suggests the potential presence of Ethereum-specific features or patterns that align more effectively with deep learning model architectures. Notably, the GRU model consistently outperforms the LSTM model across RMSE, MAE, and MAPE. These outcomes underscore the GRU model’s capacity as a robust tool for cryptocurrency value prediction. In summary, this study tackles the challenge of cryptocurrency price prediction while emphasizing the promising role of advanced neural network architectures, such as GRU, in enhancing prediction accuracy, thus offering valuable insights into financial forecasting.
Saachin Bhatt, Mustansar Ali Ghazanfar, Mohammad Hossein Amirhosseini
This research explores the impact of social media sentiments on predicting Bitcoin prices using machine learning models, integrating on-chain data, and applying a Multi Modal Fusion Model. Historical crypto market, on-chain, and Twitter data from 2014 to 2022 were used to train models including K-Nearest Neighbors, Logistic Regression, Gaussian Naive Bayes, Support Vector Machine, Extreme Gradient Boosting, and Multi Modal Fusion. Performance was compared with and without Twitter sentiment data which was analysed using the Twitter-roBERTa and VADAR models. Inclusion of sentiment data enhanced model performance, with Twitter-roBERTa-based models achieving an average accuracy score of 0.81. The best performing model was an optimised Multi Modal Fusion model using Twitter-roBERTa, with an accuracy score of 0.90. This research underscores the value of integrating social media sentiment analysis and onchain data in financial forecasting, providing a robust tool for informed decision-making in cryptocurrency trading.
Algorithmic trading enables the execution of orders using a set of rules determined by a computer program. Orders are submitted based on an asset’s expected price in the future, an approach well suited for high-volatility markets, such as those trading in cryptocurrencies. The goal of this study is to find a reliable and profitable model to predict the future direction of a crypto asset’s price based on publicly available historical data. We first develop a novel labeling scheme and map this problem into a Machine Learning classification problem. The model is then validated on three major cryptocurrencies through an extensive backtest over a bull, bear and flat market. Finally, the contribution of each feature to the classification output is analyzed.