Accurate and efficient cryptocurrency price prediction is vital for investors in the volatile crypto market. This study comprehensively evaluates nine modelsâincluding baseline, zero-shot, and deep learning architecturesâon 21 major cryptocurrencies using daily and hourly data. Our multi-dimensional evaluation assesses models based on prediction accuracy (MAE, RMSE, MAPE), speed, statistical significance (DieboldâMariano test), and economic value (Sharpe Ratio). Our research found that the optimally fine-tuned TimeGPT model (without variables) demonstrated superior performance across both Daily and Hourly datasets, with its statistical leadership confirmed by the DieboldâMariano test. Fine-tuned Chronos excelled in daily predictions, while TFT was a close second to TimeGPT for hourly forecasts. Crucially, zero-shot models like TimeGPT and Chronos were tens of times faster than traditional deep learning models, offering high accuracy with superior computational efficiency. A key finding from our economic analysis is that a modelâs effectiveness is highly dependent on market characteristics. For instance, TimeGPT with variables showed exceptional profitability in the volatile ETH market, whereas the zero-shot Chronos model was the top performer for the cyclical BTC market. This also highlights that variables have asset-specific effects with TimeGPT: improving predictions for ICP, LTC, OP, and DOT, but hindering UNI, ATOM, BCH, and ARB. Recognizing that prior research has overemphasized prediction accuracy, this study provides a more holistic and practical standard for model evaluation by integrating speed, statistical significance, and economic value. Our findings collectively underscore TimeGPTâs immense potential as a leading solution for cryptocurrency forecasting, offering a top-tier balance of accuracy and efficiency. This multi-dimensional approach provides critical, theoretical, and practical guidance for investment decisions and risk management, proving especially valuable in real-time trading scenarios.
The rapid rise of Bitcoin ignited a global frenzy over cryptocurrencies, driving a surge in their issuance and investment. Unprecedented returns on these assets have led to comparisons with irrational exuberance. Against this backdrop, and adopting a behavioural finance perspective, this thesis thoroughly investigates the role of investor sentiment in cryptocurrency prices and volatility across three empirical studies. Sentiment is proxied by the Fear and Greed Index (FGI), published by Alternative.me, and the Economic News Sentiment Index (NSI), developed by the Federal Reserve Bank of San Francisco. The first study examines volatility connectedness among six Bitcoin (BTC) currency pairs and identifies its determinants. The results indicate that BTC/USD and BTC/GBP are major volatility transmitters, while BTC/USDT remains relatively isolated. Key determinants include trading volume, sentiment, economic policy uncertainty, and gold volatility. There is an asymmetric effect of sentiment derived from the FGI: optimistic sentiment intensifies volatility connectedness, whereas pessimistic sentiment dampens it. Policy uncertainty and gold volatility are positively associated with spillover intensity, with cryptocurrency-related events further shaping the degree of connectedness. The second study focuses on intraday cross-exchange (i.e., Bitfinex and Kraken) price discovery for Bitcoin and Ethereum, along with its potential drivers. The results show that Bitfinex dominates price discovery during most periods, but the pandemic alters this process. We identify a shift in drivers, with market quality factors losing significance and news sentiment emerging as a more prominent influence in the post-pandemic period. During episodes of heightened news sentiment, Bitfinex consolidates its leading position. Additionally, Ethereumâs price discovery is significantly associated with intraday volatility. The third study assesses whether the inclusion of the FGI in GARCH(1,1), EGARCH(1,1), and HAR(1,7,30) models improves the accuracy of volatility forecasts for twelve cryptocurrencies. The findings suggest that incorporating the FGI enhances predictive performance, with variation across assets. Bitcoin, Ethereum, and cryptocurrencies technologically linked to them (e.g., Litecoin, DASH and Ethereum Classic) benefit from the integration of sentiment in most cases. EGARCH+FGI and HAR+FGI consistently outperform competing models, particularly at the weekly forecast horizon. This thesis provides behavioural finance insights for cryptocurrency investors and policymakers. Investors may leverage the FGI and the NSI to refine trading strategies and inform venue selection, while policymakers may incorporate the FGI into monitoring frameworks to better identify systemic risks and anticipate excessive cross-market spillovers.
Cryptocurrency is an alternative payment method developed with encryption techniques. To predict Bitcoin values using both weekly and monthly datasets, this study compares four machine learning models: GRU, Weighted LSTM, LSTM, and LSTM with Attention. The models' accuracy and dependability in capturing the dynamics of cryptocurrency prices were assessed using Mean Squared Error (MSE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-Squared (RSCORE). While LSTM with Attention did well with an RSCORE of 0.7173, LSTM with Attention had the highest RSCORE of 0.9173 in the weekly dataset, indicating higher ability in modelling short-term sequential patterns. Additionally, weighted LSTM performed well (RSCORE of 0.8002), surpassing GRU (RSCORE of 0.5728), which had trouble keeping up with the volatility of Bitcoin prices. Both LSTM and LSTM with Attention performed best in the monthly dataset, each with the lowest MSE (0.0304) and an RSCORE of 0.8173. With an RSCORE of 0.7002, weighted LSTM came next, using temporal weighting to enhance predictions. Because of its limited capacity to grasp intricate temporal connections, GRU continuously fared poorly in both datasets. According to the analysis, LSTM is the most dependable model for both short-term and long-term forecasts, and for weekly forecasts, LSTM with Attention provides improved interpretability. These results provide a framework for applying machine learning approaches to financial time series forecasting, highlighting the significance of choosing suitable models based on data frequency, volatility, and prediction aims.
Muhammad Ali Nawaz, Wajid Alim, Sammar Abbas, Shahid Manzoor Shah ¡ 5 authors
The study investigates the co-movement relationships between cryptocurrencies and South Asian stock markets, focusing on five leading cryptocurrencies: Bitcoin, Ethereum, Tether, Binance Coin, and Ripple, and five South Asian stock indices: BSE, PSX 100, DSE 30, NEPSE, and Sri Lanka's All Share Index, and also used five major global indices for the accuracy of analysis. The study aims to understand their integration and causal dynamics. The analysis uses 357 weekly observations of historical prices from November 6, 2017, to September 2, 2024, applying econometric tools such as the Augmented Dickey-Fuller and Phillips-Perron tests, Johansen's Cointegration Test, Vector Auto-Regression, Vector Error Correction Model, and Granger causality to examine statistical properties, integration, and causality among the variables. Results show significant cointegration and causality between cryptocurrencies and South Asian stock indices, with cryptocurrency prices exhibiting higher volatility and faster adjustments than stock indices. These findings provide actionable insights for investors, policy-makers, and researchers regarding regulation and cross-market investment strategies. This study uniquely explores the interplay between emerging digital assets and traditional finance in a South Asian context, offering novel evidence on volatility dynamics and causal relationships that inform coupled regulatory frameworks and cross-market investment planning.
Cryptocurrencies have grown as one of the significant technological advancements in todayâs digital age. Investors all around the globe have shown a growing interest in cryptocurrencies in recent times because of their potential to bring huge profits. This paper tries to determine the best cryptocurrencies for investment by evaluating six leading cryptocurrencies, selected based on market capitalization and global relevance. Cryptocurrencies are ranked using six important evaluation criteria using two multi-criteria decision-making (MCDM) techniques: Technique for Order Preference by Similarity to Ideal Solution (TOPSIS), and VIseKriterijumska Optimizacija I Kompromisno Resenje (VIKOR), employed under hyperbolic fuzzy set (HyFS) settings. In addition, a closeness coefficient is introduced in TOPSIS method to enhance decision precision and sensitivity. Although the results obtained by the two methods are very close, investors would not want to have any doubt about the investment they are considering. The rankings from each procedure are then combined using a social choice function: the Borda Count method. Combining these approaches, the findings indicate that Bitcoin and Ethereum emerge as the most favorable cryptocurrencies for investment among the selected alternatives. To ensure the consistency of the obtained rankings, statistical analyses were conducted. Spearmanâs rank correlation coefficient (Ď), Kendallâs tau (Ď), and the Wilcoxon Signed-Rank Test were utilized to measure the degree of agreement between the rankings generated by Hyperbolic Fuzzy TOPSIS and Hyperbolic Fuzzy VIKOR. Combining these approaches to create a trustworthy decision-making model for cryptocurrency selection makes this work novel.
The rapid rise of the prices of cryptocurrencies has intensified the need for robust forecasting models that can capture the irregular and volatile patterns. This study aims to forecast Bitcoin prices over a 15-day horizon by evaluating and comparing two distant predictive modeling approaches: the Bayesian State-Space model and Long Short-Term Memory (LSTM) neural networks. Historical price data from January 2024 to April 2025 is used for model training and testing. The Bayesian model provided probabilistic insights by achieving a Mean Squared Error (MSE) of 0.0000 and a Mean Absolute Error (MAE) of 0.0026 for training data. For testing data, it provided 0.0013 for MSE and 0.0307 for MAE. On the other hand, the LSTM model provided temporal dependencies and performed strongly by achieving 0.0004 for MSE, 0.0160 for MAE, 0.0212 for RMSE, 0.9924 for R2 in terms of training data and for testing data, and 0.0007 for MSE with an R2 of 0.3505. From the result, it indicates that while the LSTM model excels in training performance, the Bayesian model provides better interpretability with lower error margins in testing by highlighting the trade-offs between model accuracy and probabilistic forecasting in the cryptocurrency markets.
Tulika Shrivastava, Basem Suleiman, Sachit A. J. Desa, Muhammad Johan Alibasa ¡ 6 authors
Abstract The volatility of cryptocurrencies necessitates reliable short-term price prediction models for informed investment decisions. This work presents two benchmarking studies that predict cryptocurrency price over hourly and daily time horizons using market indicators and social media data. Study 1 used BERT-based sentiment analysis of hourly Twitter data combined with financial indicators, while Study 2 applied VADER sentiment analysis to daily Twitter and Google Trends data alongside financial indicators. Both studies systematically evaluated statistical models (ARIMA, ARIMAX), machine learning approaches (SVR), and deep learning architectures (1D-CNN, LSTM) including ensemble, multi-modal, and hybrid configurations. Particular attention was given to the influence of lag periods, data aggregation, and sentiment analysis nuances on cryptocurrency price. Empirical results identify LSTM as the best-performing singular prediction model, achieving a 64.5% reduction in RMSE (4.56e $$-$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mo>-</mml:mo> </mml:math> 03) compared with the SVR baseline in Study 1. In Study 2, the hybrid LSTM + ARIMA model delivered the strongest performance, reducing RMSE by 32.5% (RMSE=2.55e+02) relative to the best performing singular baseline. Hybrid architectures combining LSTM with ARIMA or ARIMAX consistently achieved the lowest RMSE values, outperforming all other configurations and proving especially effective at capturing price movements and turning points. These findings demonstrate how combining statistical methods with deep learning can address non-stationarity, improve sentiment preprocessing, and enhance model interpretability.
This paper proposes a cryptocurrency portfolio trading system (CPTS) that optimizes trading performance in the cryptocurrency futures market by leveraging reinforcement learning and timeframe analysis. By employing the advantage actorâcritic (A2C) algorithm and analysis of variance (ANOVA) portfolios are constructed over multiple timeframes. Data corresponding to the trade of 18 major cryptocurrencies on Binance Futuresââbetween January 2022 and December 2023ââare used to show that trading strategies can be effectively categorized into those with high-frequency (10, 30, and 60 min) and low-frequency (daily) timeframes. Empirical results demonstrate statistically significant differences in returns between these timeframe groups, with major cryptocurrencies (e.g., Bitcoin and Ethereum) exhibiting higher returns in high-frequency trading (16â17%) than in daily trading (6â7%) during training. Performance evaluation during the test period revealed that the low-frequency group achieved a 43.06% average return, significantly outperforming the high-frequency group (5.68%). The ANOVA results confirm that both the frequency type and portfolio selection significantly influence trading performance at the 5% significance level. This study offers a novel approach to cryptocurrency trading that considers the distinct characteristics of different timeframes. The effectiveness of combining reinforcement learning with statistical analysis for portfolio optimization in highly volatile cryptocurrency markets is demonstrated.
This study is the first to integrate recurrence plots, recurrence quantification analysis (RQA) and short-time Fourier Transform (STFT) to predict cryptocurrency market behaviour. Recurrence plots, RQA statistics and STFT spectrograms were calculated from return data and used as input in random forest algorithms as they are optimal tools for identifying non-linear dynamics in market data and analyse their frequency. Our optimised XGBoost algorithm provided a forecasting AUC above 76.7% and accuracy of 70% in predicting increasing or decreasing returns. This highlights the modelâs ability to support cryptocurrency investment decision-making within an interpretable machine learning framework.
Cryptocurrency markets are highly volatile and influenced by both price trends and market sentiment, making effective portfolio management challenging. This paper proposes a dynamic cryptocurrency portfolio strategy that integrates technical indicators and sentiment analysis to enhance investment decision-making. Market momentum is captured using the 14-day Relative Strength Index (RSI) and Simple Moving Average (SMA), while sentiment signals are extracted from news articles with VADER and further validated using the Google Gemini large language model. These signals are incorporated into expected return estimates and used in a constrained mean-variance optimization framework. Backtesting across multiple cryptocurrencies shows that the integrated approach outperforms traditional benchmarks, including momentum strategy, Bitcoin Long-Short strategy, and an equal-weighted portfolio, achieving stronger risk-adjusted returns and more consistent cumulative growth. Furthermore, comparing the sentiment-only and technical-only strategies shows that incorporating sentiment information alongside technical indicators can lead to more consistent performance gains. However, the strategies exhibit substantial drawdowns that coincide with known periods of market stress, indicating that additional risk-management components are required to improve stability.
Cryptocurrency markets are characterized by ex-treme volatility, making accurate forecasts essential for effective risk management and informed trading strategies. Traditional deterministic (point) forecasting methods are inadequate for capturing the full spectrum of potential volatility outcomes, underscoring the importance of probabilistic approaches. To address this limitation, this paper introduces probabilistic fore-casting methods that leverage point forecasts from a wide range of base models, including statistical (HAR, GARCH, ARFIMA) and machine learning (e.g. LASSO, SVR, MLP, Random Forest, LSTM) algorithms, to estimate conditional quantiles of cryp-tocurrency realized variance. To the best of our knowledge, this is the first study in the literature to propose and systematically evaluate probabilistic forecasts of variance in cryptocurrency markets based on predictions derived from multiple base models. Our empirical results for Bitcoin demonstrate that the Quantile Estimation through Residual Simulation (QRS) method, partic-ularly when applied to linear base models operating on log-transformed realized volatility data, consistently outperforms more sophisticated alternatives. Additionally, we highlight the robustness of the probabilistic stacking framework, providing comprehensive insights into uncertainty and risk inherent in cryptocurrency volatility forecasting. This research fills a sig-nificant gap in the literature, contributing practical probabilistic forecasting methodologies tailored specifically to cryptocurrency markets.
The cryptocurrency market, which is extremely volatile and has high price fluctuations, is transforming the financial ecosystems in the world.In contrast to traditional markets, cryptocurrencies are characterized by the unprecedented volatility due to the complicated interaction of speculative trading, regulatory changes, technological breakthroughs, and macroeconomic forces.The purpose of the current study is to build and test machine learning models to predict the price trend of cryptocurrencies, including the most popular ones, Bitcoin (BTC), Ethereum (ETH), and other top altcoins that are traded in the United States.The analysis is based on a large amount of data on historical prices at daily, hourly, and minute-by-minute intervals, including the detailed data on opening, closing, high, and low prices, and trading volumes that indicate the liquidity and the activity of investors.The most important technical indicators such as moving averages, Relative Strength Index (RSI) and Bollinger Bands are incorporated to identify the most important market signals and momentum.It uses three machine learning models, including Logistic Regression, Random Forest Classifier, and XGBoost Classifier.Directional prediction capability (upward or downward price movements) is evaluated by accuracy, precision, recall, and F1-score measures of model performance.Logistic Regression was the most accurate among the models that were tested, which highlights its comparative effectiveness in this application.The introduction of AI-based predictive analytics into cryptocurrency trading can be a great way to improve the process of decision-making by traders and institutional investors and help them comply with regulations in the U.S. financial system.This study sheds light on the transformational nature of machine learning in cryptocurrency prediction and also points out the research opportunities in the future, especially the use of deep learning models like the Long Short-Term Memory (LSTM) network in time-series analysis.
As Bitcoin continues to establish itself as a global asset and discussions around relevant regulations become more active, there is an increasing demand for a comprehensive price prediction framework. To address this necessity, this study aims to enhance the accuracy of Bitcoin price predictions by integrating sentiment information with technical indicators, on-chain data, and cryptocurrency price data. Recognizing Bitcoinâs sensitivity to market sentiment, the proposed framework incorporates sentiment features derived from both lexicon-based methods and large language models. As unsupervised sentiment tools can introduce label noise particularly in domain-specific or ambiguous financial contexts, this study combines the outputs of multiple sentiment models at the feature level to construct a more stable representation. This design improves the robustness of downstream regression performance and distinguishes the framework from previous hybrid models that relied on a single sentiment source without component-wise evaluation. Experimental results using a dataset spanning 2700 days showed that the long short-term memory (LSTM) model with a 3-day window achieves the best performance with mean absolute percentage error (MAPE) of 3.93% and R-squared value of 0.99106. Feature importance analysis further demonstrates sentiment index as the most impactful feature, as excluding it resulted in the largest decline in predictive accuracy. Additionally, the model's performance was evaluated under four major volatility periods, revealing MAPE values ranging from 1.49 to 4.03%, highlighting the frameworkâs practical capability in rapidly adapting to sudden market shifts. In summary, integrating sentiment information attained from multiple language models significantly enhanced prediction accuracy compared to single source approaches. These findings highlight the frameworkâs practical value for sentiment-informed investment strategies and risk alerts, with a modular design that enables flexible adaptation and potential integration into automated trading systems.
In recent years, computational intelligence techniques have significantly contributed to the automation and optimization of trading strategies. Despite the increasing sophistication of predictive models, classical technical indicators such as dual Simple Moving Averages (2-SMA) remain popular due to their simplicity and interpretability. This work proposes an adaptive trading system that combines the 2-SMA strategy with a learning-based metaheuristic optimizer known as the Learning-Based Linear Balancer (LB2). The objective is to dynamically adjust the strategyâs parameters to maximize returns in the highly volatile cryptocurrency market. The proposed system is evaluated through simulations using historical data of the BTCUSDT futures contract from the Binance platform, incorporating real-world trading constraints such as transaction fees. The optimization process is validated over 34 training/test splits using overlapping 60-day windows. Results show that the LB2-optimized strategy achieves an average return on investment (ROI) of 7.9% in unseen test periods, with a maximum ROI of 17.2% in the best case. Statistical analysis using the Wilcoxon Signed-Rank Test confirms that our approach significantly outperforms classical benchmarks, including Buy and Hold, Random Walk, and non-optimized 2-SMA. This study demonstrates that hybrid strategies combining classical indicators with adaptive optimization can achieve robust and consistent returns, making them a viable alternative to more complex predictive models in crypto-based financial environments.
Mansi Sharma, Enrico Sartor, Marc Cavazza, Helmut Prendinger
Forecasting stock and cryptocurrency prices is challenging due to high volatility and non-stationarity, influenced by factors like economic changes and market sentiment. Previous research shows that Echo State Networks (ESNs) can effectively model short-term stock market movements, capturing nonlinear patterns in dynamic data. To the best of our knowledge, this work is among the first to explore ESNs for cryptocurrency forecasting, especially during extreme volatility. We also conduct chaos analysis through the Lyapunov exponent in chaotic periods and show that our approach outperforms existing machine learning methods by a significant margin. Our findings are consistent with the Lyapunov exponent analysis, showing that ESNs are robust during chaotic periods and excel under high chaos compared to Boosting and NaĂŻve methods.
Cryptocurrencies have become prominent alternative investments. Unlike traditional financial assets, their intrinsic value is a subject of ongoing debate since they do not have a tangible backing asset. As a result, investor sentiment heavily influences price volatility and serves as a key indicator of perceived value based on collective investor beliefs. However, major events such as the FTX scandal can severely weaken investor confidence. Social media drives market discussions, making sentiment analysis vital for understanding behavior and predicting price movements. This study examined sentiment analysis techniques to construct an investor sentiment index and investigate its relationship with cryptocurrency returns during the FTX collapse. We employed DistilBERT and the AFINN lexicon method to develop sentiment index, finding that DistilBERT achieves an F1-score of 76.49%, significantly outperforming AFINN's 30.65%. Furthermore, our results indicate a positive correlation between investor sentiment and cryptocurrency returns during the FTX collapse. Our findings indicate that deep learning models can be more effective than lexicon-based approaches for sentiment analysis in financial markets
This study constructs a machine learning-driven multi-factor model for Ethereum quantitative trading, combining traditional technical indicators (RSI, MACD), on-chain metrics (gas usage, active addresses), and X platform social sentiment to predict short-term returns. Backtesting from Q4 2021 to Q3 2024, using online learning and genetic algorithms for dynamic factor updates, yields a 97% annualized return, a Sharpe ratio of 2.5, and an information ratio of 1.2, outperforming Ethereum's raw returns. Simulated trading in Q4 2024 (bull market) achieves a 33% quarterly return with an 18% maximum drawdown, while Q1 2025 (bear market) records a -10% quarterly return with a 12% drawdown, confirming robustness. Technical and sentiment factors drive performance, though a 22% maximum drawdown in backtesting highlights volatility risks. An optimal Z-score threshold (Âą1.0) and 4-hour trading frequency balance profitability and costs. Future enhancements include high-frequency mainnet data integration and advanced risk management to strengthen model resilience in Ethereum's volatile market.
Cryptocurrency markets, characterized by high volatility, limited regulation, and rapid digitalization, represent an area that requires in-depth analysis of investor behavior and market dynamics. In this context, herd behavior emerges as a phenomenon where investors follow collective movements instead of making individual decisions, leading to market anomalies. The dynamics of cryptocurrency markets differ from those of traditional financial markets, necessitating alternative analytical methods to understand irrational investment decisions within these markets. This study aims to investigate herd behavior in cryptocurrency markets and its impact on market volatility. It identifies that during periods of high market volatility, investors tend to exhibit more irrational movements, following herd behavior, which disturbs market equilibrium.
Margherita Renieri, Letterio Galletta, Alberto Lluch Lafuente, Aleksander Junge ¡ 5 authors
Automated Market Makers ( AMM s) are one of the most used Decentralized Finance services enabling users to exchange crypto-assets directly without intermediaries. However, current protocols impose significant constraints on the liquidity levels required for transactions. In this paper, we propose a liquidity-saving mechanism designed to minimize the liquidity required by AMM services. Our mechanism delays the transactions violating the liquidity constraints in a queue, and, when certain conditions are met, it selects from the queue a feasible transaction sequence that fulfills the constraints and executes them atomically on the blockchain. We provide an operational semantics of such a mechanism that precisely characterizes the interactions between users and AMM s and the conditions when the liquidity-saving mechanism is triggered. Moreover, we show that our mechanism allows for novel liquidity saving behavior for multi-party exchange, multi- AMM arbitrage, and enhances user intent compared to traditional AMM s. Finally, to validate our approach, we develop a simulator and experiment with various application scenarios, yielding insights into the practical implications of our mechanism.
Tingsheng Feng, Zhihao Shen, Xi Zhao, Xiaoni Lu ¡ 5 authors
Accurately predicting user trading behavior in decentralized exchanges is essential for investors to mitigate risks and optimize their trading strategies. While existing research primarily focuses on predicting trading behavior in stock markets, these methods often struggle to adapt to the distinct nature of cryptocurrency trading. Specifically, they face issues such as limited adaptivity to high-frequency and algorithmic trading, as well as an insufficient consideration of fine-grained real-time market participants' behavior.Thanks to the pending mechanism of blockchain, it becomes possible to capture traders' interactions before transactions are finalized, providing valuable insights into market state. However, accurately modeling and predicting trading behavior in decentralized exchanges presents challenges, including limited adaptability to high-frequency trading, a lack of fine-grained transaction data, and high computational costs. This work proposes CryptoMixer, a lightweight fine-grained market information-aware multilayer perceptron (MLP)-based model for high-frequency cryptocurrency trading behavior prediction. Specifically, to overcome the sparsity and asynchrony of user behavior data, CryptoMixer develops a Market Information Augmenter that aggregates historical transaction data of users. Furthermore, CryptoMixer designs a Market Information Mixer as well as a Two-stream MLP Fusion Mixer to capture fine-grained user trading behavior patterns. We evaluate CryptoMixer on real-world user trading datasets from the Uniswap decentralized finance platform. Experimental results demonstrate that CryptoMixer outperforms traditional prediction models while maintaining low computational overheads, providing a practical solution for real-time cryptocurrency trading behavior prediction. The code is available at https://github.com/aqua111000/CryptoMixer.
Ignacio Ariel Del Monte, Juan de Lucio, Miguel Angel Sicilia Urban
This systematic review examines risk of Impermanent Loss (IL) in Automated Market Makers (AMMs) within the Decentralized Finance (DeFi) ecosystem, employing the PRISMA-S methodology. Our comprehensive search across the Web of Science and Scopus databases identified 38 relevant studies published between January 2020 and September 2024. The review reveals a predominant focus on Constant Product Market Makers (CPMMs), which comprise 55.7% of all mentions, underscoring their central role in DeFi markets. There are 9 underlying causes affecting IL risk and the most important ones are price volatility, asset imbalance, and risk/return management. According to our categorization, the most commonly used Mitigation Strategies are Investment Strategies, Decentralized Tools and Technologies, and Design and Management of Liquidity Pool, Hedging Strategies and Context Strategies. IL risk research is calculated equally theoretically and empirically (9 references for each) and there are 6 research papers that calculate in a mixed way. Only 13 research papers employ market data in their reviews and 7.9% of all papers measure IL risk quantitatively. We seek to focus our studies on a more detailed treatment of the risk of IL that will result in improvements to liquidity providers in DeFi.
Cryptocurrency markets are highly volatile, creating challenges for accurate risk management and forecasting. As digital assets become more integrated into financial systems, understanding their volatility dynamics is essential for investors and policymakers. Previous research has primarily applied standard GARCH models to cryptocurrencies, often neglecting advanced specifications that capture asymmetry, regime-switching, and long-memory effects. This limits the accuracy of volatility forecasts and fails to reflect the unique behaviour of digital assets. This study aims to identify the most effective GARCH-class models for forecasting volatility in Bitcoin, Ethereum, Binance Coin, and Ripple. We analyse daily returns from August 2017 to December 2024, applying eight advanced GARCH specifications: EGARCH, GJR-GARCH, FIGARCH, HYGARCH, MSGARCH, CS-GARCH, and Log-GARCH. Hyperparameter tuning is conducted via grid search across lag orders (p, q â [1, 5]), mean equations, and error distributions. Model performance is evaluated using AIC, BIC, RMSE, and MAE. Results show that MSGARCH and EGARCH outperform symmetric and short-memory models, highlighting the importance of regime-switching and leverage effects. FIGARCH provides the best fit for Bitcoin and Ethereum, confirming long-memory persistence. Skewed Studentâs t and GED distributions improve accuracy by capturing heavy tails and asymmetry. These findings demonstrate the limitations of standard GARCH models and underscore the value of advanced specifications in modelling cryptocurrency volatility. The study offers practical insights for traders and risk managers, contributing to more robust forecasting in non-stationary markets. Advanced GARCH models significantly enhance volatility prediction for digital assets. Future research could extend this framework to other speculative instruments or integrate machine learning techniques to further improve performance.