Bitcoin, the pioneering cryptocurrency, is renowned for its extreme volatility and speculative nature, making accurate price prediction a persistent challenge for investors. While recent studies have employed multivariate models to integrate historical price data with social media sentiment analysis, this study focuses on improving an existing univariate approach By incorporating sentiment and tweet volume data into a multivariate framework, we systematically evaluated the benefits of this integration. Among the five LSTM-based models developed for this study, the Multi-LSTM-Sentiment model achieved the best performance, with the lowest mean absolute error (MAE) of 0.00196 and root-mean-square error (RMSE) of 0.00304. These results underscore the significance of including social media sentiment in predictive modelling and demonstrate its potential to enhance decision-making in the highly dynamic cryptocurrency market.
Jahanzaib Alvi, Kehkashan Nizam, S. M. A. Jafri, Muhammad Rehan ¡ 5 authors
This paper investigates the efficacy of deep learning models such as Long-Short Term Memory (LSTM), Convolutional Neural Networks (CNN), and Gated Recurrent Units (GRU) for cryptocurrency price prediction, examining their short-term and long-term forecasting accuracy for investor guidance and advancing AI in financial analysis. The study uses time series analysis with LSTM, CNN, and GRU models on daily cryptocurrency prices from Investing.com, preprocessing data before testing on Bitcoin, Ethereum Classic, Ethereum, Litecoin, Monero, and the other 37 cryptocurrencies. RMSE, MAE, and accuracy rates measure performance. Findings revealed that only six cryptocurrencies were selected for final analysis, including Bitcoin, Ethereum Classic, Ethereum, Litecoin, and Monero. Results indicate that the deep learning models, particularly the LSTM and GRU, can predict cryptocurrency prices with high accuracy, especially for short-term forecasts within a 7-day window. The CNN model demonstrates significant predictive power, suggesting its utility for immediate trading decisions. Across the models, short-term precision was remarkably high, while long-term predictions maintained a moderate level of accuracy. This study presents a comparative analysis of LSTM, GRU, and CNN models for forecasting cryptocurrency prices, emphasizing LSTM and GRU's ability to navigate price volatility and suggesting their use for real-time trading analysis. The study's historical data reliance curtails forecasting unforeseen market shifts. Future studies should include new variables like social sentiment and blockchain analytics and test real-time adaptive models to enhance predictive strength. Model validation in actual market conditions is recommended for practical application.
This research compares the effectiveness of the hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) method and the Convolutional Neural Network (CNN) method in predicting the Ethereum (ETH) exchange rate against the United States Dollar (USD). The research uses historical ETH/USD data from Yahoo Finance for the period 2017-2022. Evaluation of the two models was carried out using the performance metrics Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), coefficient of determination (R²), and accuracy rate. The results showed that the CNN-LSTM hybrid model significantly outperformed the CNN model in predicting the ETH/USD exchange rate with a Test RMSE value of 94.67 compared to 129.02 for CNN, as well as an accuracy rate of 96.31% versus 94.89%. These findings contribute to the fintech literature by providing empirical evidence of the superiority of hybrid methods for high volatility cryptocurrency exchange rate prediction.
With the rapid development of the financial industry and artificial intelligence (AI) technology, the application of AI robots in finance has become a widely discussed topic in the academia. As an important part of the financial market, the secondary trading market of virtual currencies is characterized by high volatility, risk and decentralization, which poses significant challenges for traditional trading methods. AI technologies, especially machine learning and deep learning algorithms, provides a new path to optimize trading strategies and reduce investment risks thanks to their powerful data processing, pattern recognition and real-time analysis capabilities. This paper focuses on the characteristics of AI technology and the secondary trading market of virtual currencies, as well as the practical application of artificial intelligence technology in the financial industry, and explores the potential of applying AI robots to virtual currency trading. The research shows that AI robots can provide more accurate decision support for investors through massive data analysis, automatic trading execution and real-time risk assessment, improving market response speed and investment return rate. If the combination of AI robots and virtual currency trading is successful, it is expected to create more long-term and stable returns for investors, providing important theoretical and practical value for the development of financial investment.
Zhi Zhan Lua, Chee Kiat Seow, Raymond Ching Bon Chan, Yiyu Cai ¡ 5 authors
Distributed ledger technology (DLT) and cryptocurrency have revolutionized the financial landscape and relevant applications, particularly in investment opportunities. Despite its growth, the marketâs volatility and technical complexities hinder widespread adoption. This study proposes a cryptocurrency trading system powered by advanced machine learning (ML) models to address these challenges. By leveraging random forest (RF), long short-term memory (LSTM), and bi-directional LSTM (Bi-LSTM) models, the cryptocurrency trading system is equipped with strong predictive capacity and is able to optimize trading strategies for Bitcoin. The up-to-date price prediction information obtained by the machine learning model is incorporated by custom oracle contracts and is transmitted to portfolio smart contracts. The integration of smart contracts and on-chain oracles ensures transparency and security, allowing real-time verification of portfolio management. The deployed cryptocurrency trading system performs these actions automatically without human intervention, which greatly reduces barriers to entry for ordinary users and investors. The results demonstrate the feasibility of creating a cryptocurrency trading system, with the LSTM model achieving a return on investment (ROI) of 488.74% for portfolio management during the duration of 9 December 2022 to 23 May 2024. The ROI obtained by the LSTM model is higher than the performance of Bitcoin at 234.68% and that of other benchmarking models with RF and Bi-LSTM over the same timeframe. This approach offers significant cost savings, transparent portfolio management, and a trust-free platform for investors, paving the way for broader cryptocurrency adoption. Future work will focus on enhancing prediction accuracy and achieving greater decentralization.
The cryptocurrency market, known for its high volatility and immense data availability, provides an excellent opportunity for predictive modeling. This paper explores the prediction of Ethereumâs price using four distinct models: Random Forest, Logistic Regression, Long Short-Term Memory Networks (LSTM), and CNN-LSTM hybrid models. The study evaluates the performance of these models based on metrics such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Rsquared ( R2 ), and Accuracy (%). The findings highlight that Logistic Regression outperformed the other models with the lowest MSE (6741.12) and highest accuracy (98.66% ) [Table 1]. This research demonstrates the potential of combining traditional and advanced machine learning techniques to achieve robust price prediction in the cryptocurrency domain.
The inherent challenges of financial time series forecasting demand advanced modeling techniques for reliable predictions. Effective financial time series forecasting is crucial for financial risk management and the formulation of investment decisions. The accurate prediction of stock prices is a subject of study in the domains of investing and national policy. This problem appears to be challenging due to the presence of multi-noise, nonlinearity, volatility, and the chaotic nature of stocks. This paper proposes a novel financial time series forecasting model based on the deep learning ensemble model LSTM-mTrans-MLP, which integrates the long short-term memory (LSTM) network, a modified Transformer network, and a multilayered perception (MLP). By integrating LSTM, the modified Transformer, and the MLP, the suggested model demonstrates exceptional performance in terms of forecasting capabilities, robustness, and enhanced sensitivity. Extensive experiments are conducted on multiple financial datasets, such as Bitcoin, the Shanghai Composite Index, China Unicom, CSI 300, Google, and the Amazon Stock Market. The experimental results verify the effectiveness and robustness of the proposed LSTM-mTrans-MLP network model compared with the benchmark and SOTA models, providing important inferences for investors and decision-makers.
Abstract We present the first evidence of investorâtradingâbased disagreement's influence on crossâsectional cryptocurrency daily returns. We interpret abnormal trading volume as investor disagreement and find evidence in support of Miller's disagreement model: when shortâsale constraints are binding, high abnormal volume (high disagreement) assets experience lower future returns. Further supporting Miller, these same conditions associate with higher contemporaneous order imbalance, and ex post decreases in both buying and selling activities, with the former exceeding the latter in magnitude. By contrast, the effect of high disagreement disappears after a coin's margin trading is activated. We conclude that priceâoptimism models explain the disagreementâreturns relationship when opinion divergence is likely the dominant determinant of returns.
With the rapid growth of the cryptocurrency market, researchers increasingly study the price fluctuations and market behavior of digital assets. Gold, as a traditional safe-haven asset, often shows an inverse relationship with high-risk financial assets. Recently, scholars have focused on how gold market volatility affects cryptocurrencies, exploring potential co-movement or substitution effects. This study uses Python and econometric tools, including the Vector Autoregression (VAR) model, Granger causality test, impulse response functions, and forecast error variance decomposition, to analyze the impact of gold price changes on Bitcoin and Ethereum. Using weekly closing prices from 2018 to 2024, the results show that Bitcoinâs price is positively influenced by gold futures in the short to medium term, while gold shows a negative feedback response to Bitcoinâs returns with a two-period lag. Ethereum appears more independent and less affected by gold or Bitcoin. Strong interlinkages exist between Bitcoin and Ethereum, with Bitcoin playing a dominant role in influencing Ethereumâs price. This study has improved the understanding of the connections between cryptocurrencies and traditional assets., which also provides investors with insightful information on portfolio management.
This paper analyses the relationship between Bitcoin (BTC) values and the values of older cryp tocurrencies. Daily closing price data of Bitcoin (BTC), Litecoin (LTC), Namecoin (NMC), Peercoin (PPC), Ripple (XRP), Dogecoin (DOGE), Primecoin (XPM) and Nxt (NXT) for the period from De cember 16, 2013 to December 14, 2024 was used to perform Granger causality in two ways. The results show linear and nonlinear Granger causality in both directions between most cryptocu rrencies, BTC-LTC, BTC-XRP and BTC-DOGE relationships. On the other hand, BTC does not have a causal influence on NMC, PPC and XPM in the linear model. In contrast, in the nonlinear model, BTC Granger causes NMC. These insights are crucial for understanding the complex cryptocu rrency market price dynamics, aiding investors and analysts in making more informed decisions based on historical data and predictive relationships.
Abstract Bitcoin has emerged as a highly attractive and reliable investment asset for financial managers, businesses, and economic firms due to its unique features such as high security, decentralization, and potential for increased income. Consequently, Bitcoin price prediction has become a significant topic of interest among financial and economic analysts and researchers. Forecasting in such contexts often involves uncertain conditions and limited information. Grey systems theory, which specializes in analyzing problems with small samples and insufficient information, offers a promising approach. This study aims to predict the price of Bitcoin using an advanced model of grey systems theory: the fractional multivariable grey model (FGM(1, N )). The FGM(1, N ) model stands out by incorporating external factors into its predictions. Specifically, this research utilizes the FGM(1,3) model, considering the crude oil and gold prices to forecast Bitcoin price. The results demonstrate that the FGM(1,3) model provides more accurate predictions and better performance than the FGM(1,1) model, which does not include external factors like oil and gold prices. This study highlights the significant impact of crude oil and gold price trends on Bitcoin's market and underscores the effectiveness of the multivariable fractional grey model in financial forecasting.
This thesis aims to understand the nature of Bitcoin and the characteristics of Bitcoin-related equities using established asset pricing frameworks. It involves the empirical testing of two hypotheses. The first hypothesis posits that Bitcoin returns should be priced in the cross-section of expected stock returns, with a negative risk premium. Using a sample of 5,091 U.S.-listed stocks from March 2011 to April 2024, the cross-sectional analysis indicates that the risk premium associated with Bitcoin returns is not statistically significant. This finding challenges the âdigital goldâ narrative, which implies that Bitcoin functions as a safe-haven asset. Instead, the evidence suggests that portfolios with extreme Bitcoin betas consistently yield abnormal negative future returns, revealing a non-linear, inverted U-shaped relationship between Bitcoin beta and expected stock returns. While abnormal negative returns align more closely with speculative behavior, the interpretation regarding Bitcoinâs role remains theoretically challenging, as portfolios with the lowest Bitcoin betas also exhibit abnormal negative returns. The second hypothesis examines the risk determinants of Bitcoin-related stocks. This analysis is based on a sample of 20 Bitcoin-holding firms listed in the U.S. market, covering the three-year period from January 2020 to December 2022. The results indicate that the stock returns of these firms are significantly exposed to daily Bitcoin price fluctuations, exhibiting a positive beta. Additionally, the stock returns of Bitcoin-mining firms in the sample are significantly influenced by changes in Bitcoin mining difficulty, with a negative sensitivityâan effect not observed in other types of Bitcoin-holding firms. This suggests that Bitcoin-specific risk factors beyond price fluctuations may play a role in the risk-return dynamics of Bitcoin-related equities. Furthermore, a reverse size effect is observed within this sector: Bitcoin-related firms with larger market capitalizations tend to generate higher returns compared to smaller firms. This finding holds important implications for industry practice since it challenges the conventional belief that smaller stocks typically yield higher returns.
We selected the daily trading data of BTC, SPY, DXY, GLD, and QQQ from Yahoo Finance, aiming to analyze the role of BTC in portfolios. This paper believes that BTC, as a high-risk asset, is speculative. Through correlation analysis, its returns were found to be independent of other traditional assets, proving that applying BTC to investment strategies could create arbitrage opportunities. Through various asset combinations in investment portfolio experiments, we found that the intervention of BTC could enhance the returns and optimal Sharpe ratio of the original investment portfolio, and the increase in the optimal Sharpe ratio decreased as the number of assets in the portfolio except for BTC increased. Therefore, for ordinary investors, we suggest adding 10% - 20% of BTC to a single asset. Through out-of-sample testing, we found that the investment strategy that includes BTC investment based on historical data, although it could not achieve the optimal Sharpe ratio, would have higher returns than the optimal Sharpe ratio investment portfolio without BTC intervention in the current period, considering that investors have certain risk tolerance, we believe that the effectiveness of historical investment strategies can be verified.
We propose a novel machine learning framework to quantify the effects of regulatory policies on GDP normalized Bitcoin trading volume in the United States, Russia, and Indonesia. Our panel dataset integrates Bitcoin price series, country level adoption rates (2021â2024), macroeconomic indicators, and granular policy variables. An XGBoost regression model predicts future trading volume, and SHAP values to further elucidate feature importance and interactions. Using this calibrated model, we conduct policy ablation simulations by selectively removing regulatory instruments asset classification, licensing, taxation, AML/KYC stringency, and payment bans. Results indicate jurisdiction specific sensitivities: removing AML enforcement in the United States increases volume by +71.45%, while eliminating taxation in Indonesia reduces volume by -46.90%. Comprehensive removal of all regulations yields mixed outcomes: a +26.38% increase in the United States, versus -6.27% in Russia and -47.55% in Indonesia. These findings offer quantitative insights into the trade offs faced by policymakers when designing cryptocurrency regulation.