This study reviews the advancements in AI-driven methods for predicting stock prices, tracing their evolution from traditional approaches to modern finance. The role of AI in the market extends beyond predictive systems to encompass the intersection of financial markets with emerging technologies, such as blockchain, and the potential influence of quantum computing on economic modeling. A decentralized finance system examines the application of Reinforcement Learning in financial market prediction, highlighting its potential for continuous learning from dynamic market conditions. The study discusses the development of hybrid prediction models, stock market machine learning systems, and AI-driven investment portfolio management. The potential of quantum computing enhances portfolio analysis, fraud detection, optimization, and asset valuation for complex market predictions, as well as the impact of blockchain technologies on transparency, security, and efficiency. Machine learning techniques can significantly automate data collection and purification. Financial decision-making and the application of time-series analysis techniques can be readily learned through deep reinforcement learning for stock price prediction. Deep Neural Networks and Strategic Asset Allocation can be managed by evaluating performance and portfolio using real-time market insights from AI models. Although there are numerous ethical, sentimental, regulatory, and data quality issues in market prediction, the future job market is heavily dependent on these criteria, particularly through effective risk management and fraud detection.
In this work, we utilize the blockchain transactions and financial instruments to pre-dict the Bitcoin price using machine learning. We use three models: Light Gradient Boosting Machine (LightGBM), Decision Tree Regressor and Random Forest Regressor applied on a feature set which includes lagged close prices, 14-day Simple Moving Av-erage (SMA), Relative Strength Index (RSI) and daily confirmed Bitcoin transactions. The data is temporally aligned and pre-processed to maintain temporal coherence, as well as for conversational fluency. Through the results assessment by means of RMSE MAE, MAPE and R², we can found that Random Forest model has results closer to best performance with values of: 264.81 (RMSE); 175.41(MAE); for MAPE is 0.27% and; R² equals to 0.9958. Our findings also lend strong support for the effectiveness of simul-taneously considering not only blockchain-specific market variables but also tradi-tional financial predictors towards improved model performance and generalization. Our findings underscore the importance of raw blockchain transaction data for pre-dicting cryptocurrency prices, and present a new tool for data-based decision making in decentralized finance.
Kassem Danach, Hassan Rkein, Ahmad Farroukh, Ziad E. L. Balaa ¡ 5 authors
The static and hard-coded logic of smart contracts in Decentralized Finance (DeFi) platforms significantly limits their adaptability in dynamic and volatile market environments. To address this challenge, we propose a novel hyper-heuristic driven framework that enables real-time rule optimization within smart contracts, thereby enhancing responsiveness, gas efficiency, and operational robustness. The framework features a two-layer architecture: a reinforcement learning-based high-level controller selects appropriate low-level rule heuristics from a domain-specific library based on evolving transaction contexts and on-chain data. Implemented and evaluated on Uniswap v2 and Aave v3 protocols, the system dynamically optimizes parameters such as slippage tolerance, gas usage thresholds, and loan-to-value ratios. Experimental results on real-world datasets show significant performance improvements, including a 45.6% increase in transaction success rate, 28.3% reduction in average gas consumption, and 38.4% drop in liquidation events under market stress scenarios. This research demonstrates the feasibility and advantages of embedding intelligent, adaptive decision-making mechanisms within DeFi smart contracts, opening new pathways toward autonomous, resilient, and regulation-aligned blockchain systems.
For many decades, time-series forecasting has been applied to different problems by scientists and industries. Many models have been introduced for the purpose of forecasting. These advancements have significantly improved the accuracy and reliability of predictions, especially in complex scenarios where traditional methods struggled. As data availability continues to expand, the integration of machine learning techniques is likely to further enhance forecasting capabilities across various fields. Today, hybrid techniques are gaining popularity, as they combine the advantages of different approaches to deliver improved predictive performance and more advanced visualization analytics for decision support. These hybrid approaches can provide better prediction, and at the same time, they can develop a more sophisticated set of visualization analytics for decision support. Recently, the integration of cross-entropy, fuzzy logic, and attention mechanisms in hybrid forecasting models has enhanced their ability to capture complex and uncertain patterns in financial and energy markets. In this study, we propose a hybrid ANNâLSTM deep learning model optimized with cross-entropy, fuzzy logic, and an attention mechanism to enhance the forecasting of financial and energy time series, specifically Ethereum and natural gas prices. Our models combine the feature extraction strength of ANN with the temporal learning of LSTM, while cross-entropy improves convergence, fuzzy logic handles uncertainty, and attention refines feature weighting. Since inaccurate forecasts can lead to greater estimation uncertainty and increased financial and operational risk, improving predictive reliability is essential for effective risk mitigation. These techniques prove effective not only in improving estimation accuracy but also in minimizing financial risks and supporting more informed investment decisions.
While deep learning models have demonstrated superior performance in cryptocurrency forecasting, their deployment is often hindered by a lack of interpretability and trustworthiness. To bridge this gap, this paper proposes the Cryptocurrency Counterfactual Explanation (CryptoForecastCF) model. Recognizing the inherent volatility and complex non-linear dynamics of cryptocurrency markets, we argue that understanding the sensitivity of model outputs to slight variations in historical conditions is fundamental to robust risk management. CryptoForecastCF employs a gradient-based optimization strategy to generate meaningful counterfactual explanations. Specifically, it identifies minimal modifications, defined as the optimal perturbations to historical market features such as price constrained by â1 or â2 norms, that are sufficient to steer the model's future predictions into user-specified target intervals. This approach not only elucidates the key driving factors and decision boundaries of opaque models but also equips traders and risk managers with actionable insights, enabling them to identify the specific market shifts required to navigate high-stakes scenarios and mitigate unfavorable predictive outcomes.
Cryptocurrency price forecasting has gained increasing attention due to the marketâs high volatility and structural complexity. While many recent studies have explored deep learning architectures, including attention- and transformer-based models, existing research still faces notable limitations: (i) inconsistent feature engineering choices, (ii) limited examination of hybrid machine-learning models, and (iii) a lack of transparent trading evaluation using realistic backtesting assumptions. To address these gaps, this study develops a hybrid forecasting and trading framework based on Support Vector Regression (SVR) combined with a set of rule-based technical strategies. Using four major cryptocurrencies â BTC, ETH, XRP, and LTC â from 2018 to 2020, the proposed framework integrates thirteen technical indicators with a sliding-window scheme and compares SVR against Random Forest (RF) and Long Short-Term Memory (LSTM) benchmarks. Empirical results show that SVR offers a competitive balance between predictive accuracy and computational efficiency, particularly in moderate-volatility regimes. The strategy backtesting further demonstrates that SVR-driven signals can outperform traditional technical rules under certain market conditions, although limitations remain for highly volatile assets such as Bitcoin. The study contributes to the literature by clarifying feature-design choices, evaluating SVR within a multi-asset setting, and providing reproducible code and datasets through an open-access repository.
This study investigates the existence of long-run relationships between cryptocurrency prices (Bitcoin, Ethereum) and macro-economic and macro-financial variables, addressing a gap in prior research primarily focused on short-run responses. Using time series data on these variables over the period January 2022 to December 2024, statistical co-movement tests are applied to find significant relationships. Tests conducted with monthly data reveal significant co-movement in Bitcoin and Ethereum prices and the Consumer Sentiment Index, global gold reserves (measured in ounces), the MSCI World Index, and the Producer Price Index. Additionally, weekly tests reveal co-movement between Ethereum and gold prices in calendar year 2023. These findings provide empirical evidence that Bitcoin and Ethereum increasingly reflect certain macro-financial and macro-economic conditions rather than trading independently of traditional economic forces, while evidence supporting a stable âdigital goldâ role remains sparse and episodic.
This paper examines the return connectedness between Bitcoin and stock indices of economies with high levels of cryptocurrency adoption. Such economies are predominantly emerging markets characterized by elevated inflation, poor institutional quality, and macroeconomic and political instability, creating conditions under which investors may reallocate from traditional assets to Bitcoin during episodes of increased uncertainty. To assess this linkage, we employ a TVP-VAR framework with frequency-domain decomposition. Our results indicate only modest return connectedness under normal market conditions, which intensifies during periods of market turmoil. This observed pattern, along with low correlation and the identification of Bitcoin as a net return receiver, led to testing the portfolio diversification potential of Bitcoin. The evidence indicates that Bitcoin contributes to both risk mitigation and return enhancement at low hedging costs. The effect is more pronounced for emerging-market portfolios than for developed markets.
En menos de un aĂąo, Bolivia ha pasado de prohibir el uso de criptomonedas a incorporar su uso de manera frecuente. Esta investigaciĂłn examina la dinĂĄmica de los retornos y el riesgo asociado al Bitcoin, la criptomoneda de mayor valor en el ecosistema cripto, mediante modelos diseĂąados para activos de alta volatilidad. El anĂĄlisis se basa en una serie temporal de datos diarios recopilados durante doce aĂąos, con ĂŠnfasis en la mediciĂłn de retornos negativos. Los resultados muestran que la media de los retornos es positiva y estadĂsticamente significativa, aunque su capacidad explicativa sobre la variabilidad total es limitada, lo cual es consistente con el comportamiento tĂpico de series financieras de alta frecuencia. En cuanto a la volatilidad, se confirma la presencia de heterocedasticidad condicional, con efectos ARCH y GARCH altamente significativos. La persistencia de la volatilidad, evidenciada por un coeficiente GARCH cercano a uno, indica que los episodios de alta o baja volatilidad tienden a mantenerse en el tiempo. Estos hallazgos destacan la relevancia de modelar adecuadamente la varianza condicional en el anĂĄlisis de activos financieros como el Bitcoin. Adicionalmente, se identificĂł la necesidad de ajustar la escala de los datos, recomendĂĄndose una rescalaciĂłn previa para mejorar la precisiĂłn en futuras estimaciones.
The rapid expansion of decentralized finance (DeFi) has generated rich, transparent on-chain data that remains largely underutilized in high-frequency trading models. Most existing studies rely primarily on centralized exchange (CEX) price feeds, which often suffer from low signal-to-noise ratios (Lim et al., 2021; Lee et al., 2025). This study develops a multi-modal forecasting framework for the WETH/USDC 0.05% fee tier pool on Uniswap v3. We integrate Binance CEX market microstructure data with granular onchain DEX metrics, including swap imbalance, on-chain volume, liquidity depth, tick velocity, and pool liquidity utilization. An XGBoost classifier was trained on synchronized 15-minute bars, with realistic cost-aware backtesting incorporating pool fees and slippage. The model achieved a directional accuracy of 63.26% on out-of-sample data. Feature importance analysis revealed that on-chain variables, particularly volume_usd, liquidity_usd, imbalance, and tick_velocity, ranked among the top predictors. In cost-aware backtesting, the strategy outperformed Buy & Hold by 2.49 percentage points, although absolute returns remained modestly negative due to transaction costs. This research demonstrates the incremental predictive value of integrating Uniswap v3 on-chain DEX metrics with CEX data for high-frequency forecasting. While transaction costs remain a significant challenge, the findings highlight the potential of multi-modal approaches in DeFi markets and provide a foundation for future work using more advanced architectures such as the Temporal Fusion Transformer..
This thesis investigates whether the integration of digital assets, specifically Bitcoin and <br/>Ethereum, into a traditional multi-asset institutional portfolio can produce superior out-of<br/>sample risk-adjusted performance relative to the conventional 60/40 equity-bond benchmark. <br/>Using a Mean-Variance Optimization (MVO) framework enhanced by Ledoit-Wolf <br/>covariance shrinkage, the study evaluates seven distinct portfolio configurations across an <br/>empirical window spanning January 2016 to May 2026, covering multiple market regimes <br/>including the 2020 COVID-19 liquidity shock, the 2022 inflationary spike, and the 2024 <br/>institutional crypto adoption phase. <br/>The empirical design employs a dual-mode out-of-sample validation strategy: a fixed-weight <br/>stability test and a rolling realism test with systematic rebalancing. Results suggest that <br/>portfolios enriched with Bitcoin tended to achieve superior Sharpe and Sortino ratios relative <br/>to the baseline within this sample period and asset universe. The Master Portfolio (Case 7), <br/>combining equities, bonds, Bitcoin, Ethereum, Gold, and Silver, attains a static Sharpe Ratio <br/>of 1.62, a static Sortino Ratio of 3.23 and a Maximum Drawdown of only â6.40%. These <br/>findings are consistent with the primary hypothesis (Hâ) that, within this empirical setting, <br/>Bitcoin can improve out-of-sample risk-adjusted performance, while also providing evidence <br/>consistent with the secondary hypothesis (Hâ) that Bitcoin acts primarily as a portfolio <br/>diversifier rather than a consistent safe haven. All results should be interpreted as conditional <br/>on the sample period, the chosen asset universe, and the rebalancing assumptions. All asset <br/>price data was sourced from Investing.com and the risk-free rate from the FRED 3-Month <br/>Treasury Bill series (TB3MS).
The rapid expansion of the digital economy has exposed significant limitations in traditional economic frameworks, which struggle to explain phenomena such as algorithmic decisionmaking, data-driven value creation, and platform-based concentration. Existing approachesranging from production function extensions to platform models-remain fragmented and lack a unified micro-foundation. This paper proposes a behavior-centered framework to characterize economic forms and introduces the concept of economic morphology defined along four dimensions: agent structure, factor composition, behavioral pathways, and spatial distribution. Building on this framework, we define the Information Process Ratio (IPR) as a measurable indicator capturing the proportion of information-processing activities within economic behavior. Using IPR as a discriminant variable, we identify four major economic forms in human historyagricultural (IPR 10-20%), industrial (30-40%), service (50-60%), and digital (75-90%+). We show that the digital economy represents a distinct morphology, not a continuation of the industrial paradigm. Contemporary financial technology (FinTech) systems-high-frequency trading (HFT), decentralized finance (DeFi), and automated market makers (AMMs)represent extreme high-IPR regimes (95-99%), making them natural laboratories for testing the framework's predictions. We operationalize IPR using transaction-level proxies such as order-to-trade ratios (OTR), cancellation rates, and algorithmic trading share, enabling empirical application in financial markets. The framework generates testable implications linking IPR to transaction intensity, market concentration, returns to scale, algorithmic mediation, and high-frequency volatility. We further introduce the concept of IPR arbitrage, whereby economic activity flows toward higher-IPR systems, and propose a Financial Tension Index (FTI) to capture systemic strain in high-IPR environments. By shifting the analytical focus from agents to behaviors, this paper provides a unifying perspective for understanding the structural transformation of the digital economy and offers concrete implications for financial technology regulation, algorithmic market design, and systemic-risk monitoring.
Subramanya V. Odeyar, P. K. Lolakshi, L. Swetha, K. M. Thejaswini ¡ 6 authors
Abstract The Bitcoin has recently garnered significant media and public attention due to its dramatic price increases and declines. As a result, many researchers have examined the various factors influencing Bitcoinâs price and the patterns behind its fluctuations, often using machine learning techniques. This study explores several machine learning algorithms for Bitcoin price prediction, including logistic regression and long short-term memory (LSTM) models. While LSTM-based models have shown superior performance in predicting Bitcoin prices (regression), this research provides a detailed investigation into Bitcoinâs evolution and a comprehensive review of the machine learning methods used for price prediction. Additionally, the study includes a Bitcoin price prediction model, which is developed using specific algorithms to forecast Bitcoinâs price, along with insights into the factors affecting its price movements. The proposed LSTM model has achieved 98% accuracy.
Ammar Ahmed Othman, Seddiq Hassan Al-Banna Ali, Mohammed Bakr Youssef
In the digital currency, Bitcoin (BTC) is called the gold of the digital currency. It is possible to make some profits in trading of bitcoins, though this market is a very illiquid market and it is very challenging to determine the price of a bitcoin. The current work uses historical data and technical indicators to predict Bitcoin prices in a broad approach. BTC-USD price data were obtained using Yahoo Finance API and covered from 01/01/2015 till 07/01/2024. The concept of feature engineering was applied to improve the dataset by including vital financial characteristics, including Moving Averages, RSI, and Bollinger Bands for higher forecasting precision. The forward-looking model for the Bitcoin price was developed using machine learning and deep learning algorithms. The efficiency of the model was assessed with the help of Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE). The overall values of Mean Absolute Error, Mean Squared Error, and Root Mean Squared Error were 0.0062, 8.39e-05 and 0.0092 respectively which suggest that the proposed model is accurate in forecasting the future prices.
This study develops and tests a theoretical framework linking liquidity, market integration, and return predictability in cryptocurrency markets. Analyzing high-frequency daily data for five major cryptocurrencies (Bitcoin, Ethereum, Binance Coin, Solana, and Ripple), we employ Random Forest models augmented with rigorous time-series diagnostics and economic significance tests. Our findings establish a fundamental dichotomy: Bitcoin exhibits predictability driven by macroeconomic fundamentals (global risk-free rates, economic policy uncertainty), consistent with its emergence as a macro-asset; altcoins, in contrast, are dominated by internal microstructure (realized volatility, illiquidity) and speculative sentiment. Formal hypothesis tests confirm that (i) lower liquidity predicts higher future returns, (ii) machine learning models systematically underpredict during positively skewed regimes, and (iii) macro integration attenuates microstructure-driven predictability. Out-of-sample R 2 values reach 6.8% for Bitcoin and 9.6% for Ethereum, with Diebold-Mariano statistics rejecting equal predictive accuracy against a random walk at the 1% level. A mean-variance investor would earn a certainty equivalent gain of 2.4% annually by exploiting these forecasts. The results challenge the efficient market hypothesis for 1 digital assets, establish a new taxonomy of cryptocurrency predictability, and provide critical implications for asset pricing, portfolio allocation, and risk management in decentralized finance.
Predicting short-term mid-price movements from limit order book (LOB) data is a fundamental problem in quantitative finance and market microstructure research, with direct applicability to both traditional exchanges and cryptocurrency marketsâincluding centralized exchanges (CEXs) and emerging on-chain LOB protocols in decentralized finance (DeFi). We present three contributions to this domain. First, we propose DA-BiGRU-CNN, a domain-aware dual-branch architecture that decomposes LOB features into price and volume information channels, processes them through dedicated bidirectional GRU encoders with shared microstructure features, and fuses temporal representations via a multi-scale convolutional bottleneck (Conv1d with kernels k = 3,5,7). Second, we provide empirical evidence for a "feature sufficiency" hypothesis: a unidirectional GRU trained on 53 basic features achieves performance statistically equivalent to one trained on 219 extensively engineered featuresâincluding rolling statistics, exponential moving averages, and lag/difference featuresâsuggesting that recurrent hidden states implicitly learn these temporal patterns. Third, we document a "negative ensemble effect" where combining sequential (GRU) and tabular (gradient boosting) models consistently degrades prediction quality, contradicting the widely-held assumption that model diversity improves ensemble performance. On a large-scale dataset of 12,165 LOB sequences (12.1M timesteps), our GRU baseline achieves a weighted Pearson correlation of 0.266, outperforming LightGBM by 58%, while our domain-aware architecture offers an architecturally principled alternative that naturally separates price dynamics from liquidity dynamics.
Predicting cryptocurrency crashes is challenging due to their speculative dynamics, extreme volatility,and limited regulatory structure. This study investigates the predictability of Bitcoin crashesusing standard and dynamic probit models applied to daily data from 2017 to 2023. Crash indicatorsare constructed using a 3.09đ tail event rule, and model performance is evaluated through anextensive grid search over multiple crash horizons and consolidation windows, ensuring robustnessagainst horizon dependent distortions. The empirical results show that hybrid models that combinesentiment, macro-financial variables, and Bitcoin specific returns consistently outperformsentimentonly specifications.The findings highlight the importance of combining behavioral indicators with global riskmeasures and cryptocurrency specific dynamics to capture the multifaceted drivers of Bitcoincrashes. The results have implications for investors, exchanges, and policymakers looking for earlywarning mechanisms for systemic risk in digital asset markets.
This study examines the weak-form efficiency and international price integration of Malaysiaâs regulated Bitcoin market. Daily closing prices for Bitcoin traded in Malaysian ringgit (BTC/MYR), the international Bitcoin price in US dollars (BTC/USD), and the USD/MYR exchange rate are analysed over the 2021â2026 period using secondary market data. The international Bitcoin price is converted into ringgit to provide a currency-consistent benchmark for the local market. Random-walk behaviour is evaluated using the runs test, LjungâBox test and variance-ratio test. Market integration is examined through unit-root tests, EngleâGranger cointegration analysis and an error-correction model. The daily results provide mixed evidence regarding weak-form efficiency. Although the runs test does not reject randomness in return signs, the LjungâBox and variance-ratio results indicate dependence at selected horizons. This dependence becomes weaker in the weekly analysis, suggesting that the efficiency assessment is sensitive to data frequency. The local and international Bitcoin prices are cointegrated, with a long-run coefficient close to unity. The error-correction results further show that deviations from the long-run relationship are corrected over time and that international Bitcoin returns significantly influence short-run local price movements. Nevertheless, a small local price premium and residual volatility clustering remain. Overall, Malaysiaâs Bitcoin market is closely integrated with the international market but is not perfectly efficient at all horizons. The findings support policies promoting transparent benchmark pricing, market surveillance, adequate liquidity and volatility-risk controls among Malaysian digital asset exchanges.
Bitcoin price prediction is a popular topic in finance and technology circles. Developing an accurate bitcoin price prediction algorithm is crucial for the cryptocurrency market's growth and development. The development of bitcoin price prediction algorithms is challenging because bitcoin prices fluctuate heavily. Many researchers have attempted to predict the future price of Bitcoin using a variety of methods. This paper presents a web-based application for Bitcoin price prediction using a Hybrid LSTM-GRU model. The experimental results show the results of the Hybrid LSTM-GRU model compared to other models, such as LSTM and GRU. The models were evaluated using various metrics such as mean absolute error, root mean squared error, and mean squared error. The findings indicated that our model outperformed other deep learning models with RMSE, MSE, and MAE values of 0.136, 0.018, and 0.105, respectively. A web application was built using the Streamlit library.
This study investigates the relationship between two of the most important assets in the modern market, Bitcoin and gold. While gold has historically been considered a safe-haven asset, Bitcoin has emerged as a new digital alternative to gold. Using daily data from 2014 to 2025, the study applies a purely quantitative comprehensive mathematical framework that includes risk and return analysis, correlation analysis, regression models, granger causality tests, cointegration analysis, vector autoregression and impulse response functions. The results indicate that Bitcoin is a poor alternative to gold for central banks and hedgers, and a successful substitute for gold for speculators and investors, providing higher returns at a higher risk. Furthermore, even though correlation is low, investors may achieve substantial return by focusing on short term market shocks. One of the important results also include that a significant change in Bitcoin price may influence the price of Gold, but changes in the latter do not impact the former. In conclusion, the study recommends the use of Bitcoin and Gold as complements, not as substitutes.