Blockchain Papers

Arthur G. Bubolz, Giancarlo Lucca, Lizandro de Souza Oliveira, Thiago Teixeira · 7 authors

No abstract is available for this record.

Elnaz Radmand, Jamshid Pirgazi, Ali Ghanbari Sorkhi

In the digital currency market, including Bitcoin, price prediction using artificial intelligence (AI) and machine learning (ML) is critical but challenging. Conventional methods such as technical analysis (based on historical market data) and fundamental analysis (based on economic variables) suffer from data noise, processing delays, and insufficient data. To make predictions more accurate, faster, and able to handle more data, the suggested method combines several steps: extracting important information, labeling it, choosing the best features, merging different models, and fine‐tuning the model settings. Based on the price data, this approach initially generates 5 labels with a new labeling method based on the percentage of average price changes in several days and generates signals (hold, buy, sell, strong sell, and strong buy). Thereafter, it extracts 768 features from technical studies using the TA‐Lib library and from an authoritative site. The TLBOA algorithm, which does not get stuck in the local optimum with two updates, was used to select and reduce features to 15 to avoid overfitting. A variety of ML models, including support vector machine and Naive Bayes, use these selected features for training. By using the evolutionary DE algorithm to optimize the XGBoost meta‐parameters, we increased the accuracy by 1%–4%. The proposed strategy has performed better than other models, such as XGBoost with 85.66% and gradient boosting with 84.15%, and has achieved an accuracy of 91%–92%.

Syrine Ben Romdhane, Fahmi Ben Rejab, Khadija Mnasri

No abstract is available for this record.

Varun Bodepudi, Purna Chandra Rao Chinta

The increasing adoption of Bitcoin as a digital asset has led to significant interest in accurately predicting its price movements. However, the highly volatile and speculative nature of Bitcoin presents substantial challenges for traditional financial models, which often struggle to capture the complex and nonlinear patterns that influence its price fluctuations. This study proposes a novel approach to enhancing financial predictions related to Bitcoin prices by leveraging the power of big data analytics and deep learning techniques. The integration of large-scale historical market data, social sentiment analysis, blockchain transaction metrics, and macroeconomic indicators allows for a more comprehensive understanding of Bitcoin’s market behavior.To achieve this, deep learning architectures such as Long Short-Term Memory (LSTM) networks and Transformer-based models are employed due to their superior ability to capture long-range dependencies and dynamic trends in time-series data. These models are trained on high-frequency trading data, order book information, real-time market indicators, and sentiment data derived from news sources and social media platforms. By utilizing a data-driven approach, the proposed model aims to improve the robustness and accuracy of Bitcoin price predictions.Extensive experiments and comparative analyses are conducted to evaluate the effectiveness of the deep learning-based framework against traditional statistical models and classical machine learning techniques. The results demonstrate that the proposed approach significantly outperforms conventional methods in terms of predictive accuracy, stability, and generalization capabilities. The findings highlight the potential of deep learning and big data analytics in enhancing cryptocurrency market predictions and risk assessment strategies.The insights derived from this study provide valuable implications for traders, investors, and policymakers seeking to develop more informed trading strategies and risk management frameworks. By harnessing the power of deep learning and big data, this research contributes to the growing field of financial technology and underscores the importance of advanced predictive models in navigating the rapidly evolving cryptocurrency market.

Surinder Singh Khurana, Parvinder Singh, Naresh Kumar Garg

No abstract is available for this record.

Kia Jahanbin, Mohammad Ali Zare Chahooki

The impact of sentiment analysis of comments on social networks such as X (Twitter) on the cryptocurrency market’s behavior has been proven. Also, traditional sentiment analysis and not considering the possible aspects of tweets can cause the deep model to be misleading in predicting the price trend of cryptocurrencies. In this research, a model using transfer learning and the combination of pretrained DistilBERT networks, BiGRU deep neural network, and attention layer is presented to analyze the sentiments based on the aspect of tweets and predict the price trend of eight cryptocurrencies. These tweets are the opinions of 70 cryptocurrency expert influencers. After preprocessing, these tweets are injected into the hybrid model of DistilBERT, BiGRU, and attention layer (HDBA) to extract the aspect and determine the polarity of each aspect. The output of the HDBA model is entered into the combined model of BiGRU and the attention layer (HBA) to predict the price trend of each cryptocurrency in intervals of 1–10 days. The output of the HBA model is the best time interval of the influence of the sentiments of tweets on the price trend of cryptocurrencies. The results show that the HDBA model has improved the performance of the aspect‐based sentiment analysis task by an average of 3% in the benchmark datasets. The results of the HBA model also show that this model has been able to predict the best time frame of the impact of sentiments on the behavior of the cryptocurrency market with an average accuracy of 68% and a precision of 73%.

Lucas Mussoi Almeida

This dissertation presents an empirical analysis of decentralized finance through three distinct studies. By harnessing the power of on-chain data, this research delves into the mechanics of DeFi, exploring how we assess financial risk and measure market efficiency. Furthermore, it directly addresses the significant economic exter nalities of the sector by measuring the annualized energy draw of Bitcoin’s global mining industry. The first article, Risk forecasting comparisons in decentralized fi nance: An approach in constant product market makers (this research was presented at the Annual Conference of the Banco Central do Brasil (2024) and subsequently published in the Journal of Economics and Business, Volume 133, 2025; 2024 JCR Impact Factor: 3.4), pioneers by comparing risk measures between centralized and decentralized exchanges. By employing a vast dataset from Uniswap V2 Liquid ity Pool (LP) and conducting a meticulous comparative analysis of Value-at-Risk (VaR) and Expected Shortfall (ES) forecasts, using both parametric Generalized Autoregressive Conditional Heteroskedasticity (GARCH) models and the non para metric DeepAR neural network, it demonstrates that liquidity provision generally presents a statistically significant lower risk profile than an equivalent buy and hold strategy. A critical exception exists for stablecoin pairs, where protocol fees become the primary risk driver. This research provides a crucial empirical foundation for developing sophisticated, model informed risk management tools in Decentralized Finance (DeFi). The second article, Pricing efficiency in cryptocurrencies: the case of centralized and decentralized markets (published in the Journal of Economics and Business, Volume 133, 2025; 2024 JCR Impact Factor: 3.4), offers a comparative analysis of market efficiency between liquidity pool mechanisms and traditional order book systems. Utilizing Asymmetric Multifractal Detrended Fluctuation Analysis (asym metric MF-DFA) and the Thermal Optimal Path (TOP) method on data from Binance and Uniswap V2, it reveals that algorithmic LP can achieve superior weak form market efficiency compared to Centralized Exchanges (CEX) order books. The study conclusively identifies the Decentralized Exchanges (DEX) as the lead market in price discovery, transmitting signals to its centralized counterpart with an average lag of under 24 hours. This efficiency premium, driven by radical transparency and high velocity arbitrage, intensified significantly following the Ethereum 2.0 upgrade. The third article, Bitcoins halving events and the fractal nature of mining energy consumption, investigates the long term impact of Bitcoins programmed monetary policy on its mining energy consumption behavior. Applying asymmetric MF-DFA to data from the Cambridge Centre for Alternative Finance, it uncovers the com plex, multifractal nature of Bitcoins energy dynamics, showing an evolution from persistent, heterogeneous behavior before halving events toward more efficient and random consumption characteristics after each subsequent halving. The analysis provides the first documented evidence of a significant cross-chain effect, showing that Ethereum’s transition to Proof of Stake (PoS) consensus triggered an immediate and sustained decrease in the persistence of Bitcoin’s energy consumption patterns. This finding reveals previously unrecognized interconnectivity between seemingly independent networks and creates new pathways for assessing the environmental relationships within blockchain ecosystems.

Lei Shang

In this study, I propose a method for forecasting the next-day Bitcoin price range using a CART decision tree model, which integrates 124 high-dimensional technical indicators with Twitter-roBERTa sentiment analysis as the 125th feature to enhance prediction accuracy. The experiments utilize Bitcoin market data from the past six years (2019 to 2024) and approximately 58 million Twitter posts. The results demonstrate that the enhanced model, incorporating sentiment analysis, improves the average accuracy from 0.56 in the baseline model—trained solely on 124 technical indicators—to 0.62, with win rates increasing significantly by up to 45%. Sensitivity analysis further optimizes the sentiment feature weight, confirming the model’s robustness, and provides an innovative perspective for cryptocurrency market prediction, with future applications extensible through multi-source data fusion.

Aadi Singhi

This paper presents a Multi Agent Bitcoin Trading system that utilizes Large Language Models (LLMs) for alpha generation and portfolio management in the cryptocurrencies market. Unlike equities, cryptocurrencies exhibit extreme volatility and are heavily influenced by rapidly shifting market sentiments and regulatory announcements, making them difficult to model using static regression models or neural networks trained solely on historical data. The proposed framework overcomes this by structuring LLMs into specialised agents for technical analysis, sentiment evaluation, decision-making, and performance reflection. The agents improve over time via a novel verbal feedback mechanism where a Reflect agent provides daily and weekly natural-language critiques of trading decisions. These textual evaluations are then injected into future prompts of the agents, allowing them to adjust allocation logic without weight updates or finetuning. Back-testing on Bitcoin price data from July 2024 to April 2025 shows consistent outperformance across market regimes: the Quantitative agent delivered over 30\% higher returns in bullish phases and 15\% overall gains versus buy-and-hold, while the sentiment-driven agent turned sideways markets from a small loss into a gain of over 100\%. Adding weekly feedback further improved total performance by 31\% and reduced bearish losses by 10\%. The results demonstrate that verbal feedback represents a new, scalable, and low-cost approach of tuning LLMs for financial goals.

Filip Stefaniuk, Robert Ślepaczuk

The article investigates the usage of Informer architecture for building automated trading strategies for high frequency Bitcoin data. Three strategies using Informer model with different loss functions: Root Mean Squared Error (RMSE), Generalized Mean Absolute Directional Loss (GMADL) and Quantile loss, are proposed and evaluated against the Buy and Hold benchmark and two benchmark strategies based on technical indicators. The evaluation is conducted using data of various frequencies: 5 minute, 15 minute, and 30 minute intervals, over the 6 different periods. Although the Informer-based model with Quantile loss did not outperform the benchmark, two other models achieved better results. The performance of the model using RMSE loss worsens when used with higher frequency data while the model that uses novel GMADL loss function is benefiting from higher frequency data and when trained on 5 minute interval it beat all the other strategies on most of the testing periods. The primary contribution of this study is the application and assessment of the RMSE, GMADL, and Quantile loss functions with the Informer model to forecast future returns, subsequently using these forecasts to develop automated trading strategies. The research provides evidence that employing an Informer model trained with the GMADL loss function can result in superior trading outcomes compared to the buy-and-hold approach.

Gerasimos Vonitsanos, Andreas Kanavos, Phivos Mylonas

No abstract is available for this record.

Andry Alamsyah, Raras Fitriyani Astuti

Purpose This study aims to analyze public discourse on decentralized finance (DeFi) and central bank digital currencies (CBDC) using advanced natural language processing (NLP) techniques to uncover key insights that can guide financial policy and innovation. This research seeks to fill the gap in the existing literature by applying state-of-the-art NLP models like BERT and RoBERTa to understand the evolving online discourse around DeFi and CBDC. Design/methodology/approach This study uses a multilabel classification using BERT and RoBERTa models alongside BERTopic for topic modeling. Data is collected from social media platforms, including Twitter and LinkedIn, as well as relevant documents, to analyze public sentiment and discourse. Model performance is evaluated based on accuracy, precision, recall and F1-scores. Findings RoBERTa outperforms BERT in classification accuracy and precision across all metrics, making it more effective in categorizing public discourse on DeFi and CBDC. BERTopic identifies five key topics frequently discussed, such as financial inclusion, competition and growth in DeFi, with important implications for policymakers. Practical implications The insights derived from this study provide valuable information for financial regulators and policymakers to develop more informed, data-driven strategies for implementing and regulating DeFi and CBDC. Public discourse analysis enables policymakers to understand emerging concerns and trends critical for crafting effective financial policies. Originality/value This study is among the first to use advanced NLP models, including RoBERTa and BERTopic, to analyze public discourse on DeFi and CBDC. It offers novel insights into the potential challenges and opportunities these innovations present. It contributes to the growing body of research on the intersection of digital financial technologies and public sentiment.

Giovanni Arroyo, Lawrence Millen

No abstract is available for this record.

Kapil Hande, Meenakshi Chandak

No abstract is available for this record.

Seyed Alireza Athari, Derviş Kırıkkaleli, Chafic Saliba, Victoria Olushola Olanrewaju

In recent years, cryptocurrencies have emerged as a prime digital currency and an important asset, and the financial system is emerging as an important aspect while artificial intelligence (AI) has advanced expeditiously. Although AI and Bitcoin are among the most important topics in the world, empirical findings in this area are very limited. Thus, this study aims to explore co-movement between AI and Bitcoin price using quantile-based approaches from 2012 to 2024. Remarkably, the low-to-mid quantiles of AI (0.15–0.50) and the mid-to-high quantiles of BITCOIN (0.30–0.80) show a continuously positive and substantial effect from BITCOIN on AI. When AI is in its low-to-mid quantiles (0.15–0.60), it has a large and favorable impact on BITCOIN, particularly in the mid-to-upper quantiles (0.35–0.95). The results are robust by Moment Quantile Regression and Quantile-on-Quantile KRLS methods. Based on these findings policies are suggested.

Mustafa Yalçın

No abstract is available for this record.

R. Aarthi, P. Vanitha, S Reshma, C Mounisha · 5 authors

Bitcoin (BTC) and Ethereum (ETH) price and trends prediction is performed by long short-term memory (LSTM) networks, gated recurrent unit (GRU) and Random Forest machine learning algorithm, the authors explain. Feature selection techniques were effectively and widely adopted to preprocess and feed real cryptocurrency market data as input data. LSTM performs have an accuracy of 96%, GRU performs have accuracy of 97%, and Random forest 98%, meaning they are satisfactory in predicting cryptocurrency trends theme. These models were used to construct two real worlds advert based knowledge driven investment strategies which were simulated through the period under study and show the potential of this class of models. Results of which showed across different times period cases how well your prediction works [7], and all pointed out on the huge probably availability of the presence of profit making opportunity and hence the way in which your predictive way of prediction the unpredictable market crypto currency.

Nebojša Bačanin, Luka Jovanović, Miloš Mravik, Miodrag Živković · 7 authors

No abstract is available for this record.

Ch. V. Raghavendran, K. Chandra Mouli, Manu Hajari, A. Anil Kumar Reddy · 6 authors

Predictive modeling has emerged as a key focus for cryptocurrency market asset valuation due to its complex nature and high market volatility. The research looks into Ethereum price forecasting with the methods of autoregressive integrated moving average (ARIMA) and Facebook Prophet model and long short‐term memory (LSTM) networks. These models operate on historical Ethereum prices and show their efficiency regarding temporal pattern recognition and prediction accuracy. The ARIMA model helps reveal trends as well as seasonal patterns and irregularities within Ethereum price fluctuations. The Facebook Prophet model serves as a forecasting tool because it automatically handles peculiarities present within cryptocurrency price data. Time series forecasting with LSTMs becomes an advanced technique used to detect intricate patterns along with sustained dependency relationships between data points. The systematic process of preparing data and constructing models and assessing results enables proper utilization of LSTMs for predicting time series data with accuracy. Ethereum price datasets are applied to train the models which undergo performance evaluation using MPE alongside MAPE and RMSE along with MAE to reveal strengths and weaknesses during Ethereum price predictions. The evaluation shows that ARIMA and Facebook Prophet together with LSTM demonstrate success in modeling Ethereum price fluctuations. This research explores the effectiveness of time series forecasting methods for cryptocurrency price prediction yielding vital knowledge about reliable tools for financial market trend modeling. Current research findings will provide knowledge to investors and risk management professionals making decisions within the volatile digital asset space.

Lennart Ante

This paper investigates the intersection of artificial intelligence (AI) agents—autonomous software entities capable of adapting, learning, and executing multi-step operations—and decentralized finance (DeFi) ecosystems. It highlights how the adaptive decision-making capabilities, flexible governance frameworks, and data-driven optimization strategies of AI agents reshape market coordination and organizational architectures. Drawing on a qualitative analysis of 306 major crypto AI agents, the study introduces a typology that maps their diverse application areas, including algorithmic trading, portfolio management, sentiment-driven communities, and immersive entertainment. To further conceptualize the role of AI in decentralized governance, the paper develops a quadrant-based framework that distinguishes four archetypal system configurations: Traditional Decentralized Autonomous Organization (DAO) Tools, Maximally Distributed Agency, Closed Systems, and AI Dictatorships. These configurations, defined by varying degrees of autonomy and decentralization, reveal critical trade-offs between transparency, efficiency, adaptability, and control. This framework serves as a lens to theorize how AI agents reconfigure trust mechanisms, power dynamics, and decision-making processes in decentralized ecosystems. Grounded in economic and socio-technical theory, the paper positions AI agents as transformative intermediaries in tokenized environments. While demonstrating their capacity to streamline operations, enhance decision quality, and enrich user engagement, the study also addresses the governance risks posed by algorithmic control and systemic opacity. Taken together, the conceptual and empirical insights lay a foundation for ongoing interdisciplinary inquiry into the evolving role of AI agents in decentralized finance. • Introduces a typology of 306 AI agents across key DeFi application areas • Maps AI agent roles in trading, governance, community, and entertainment • Develops a governance framework for AI agent autonomy and decentralization • Shows how AI agents reduce transaction costs and reshape market structures • Highlights risks of opacity, misalignment, and centralization in DeFi AI use

Marco Corazza, Giovanni Fasano

No abstract is available for this record.

Phumudzo Lloyd Seabe, Edson Pindza, Claude Rodrigue Bambe Moutsinga, Maggie Aphane

This study presents a novel methodology for multi-step Bitcoin (BTC) price prediction by combining advanced stacking-based architectures with temporal attention mechanisms. The proposed Temporal Attention-Enhanced Stacking Network (TAESN) integrates the complementary strengths of diverse machine learning algorithms while emphasizing critical temporal features, leading to substantial improvements in forecasting accuracy over traditional methods. Comprehensive experimentation and robust evaluation validate the superior performance of TAESN across various BTC prediction horizons. Additionally, the model not only demonstrates enhanced predictive accuracy but also offers interpretable insights into the temporal dynamics underlying cryptocurrency markets, contributing to both practical forecasting applications and theoretical understanding of market behavior.

Authors unavailable

Stock price prediction is a challenging research topic because of non-linearity, significant noise and volatility of time series data.Deep learning techniques enable to learn complex and non-linear patterns of sequential time series data.Long Short-Term Memory (LSTM) is a technique which is designed to handle time series data.While LSTM model is used to extract temporal dependencies of stock data, the performance can be limited by noisy data and the challenge of capturing intricate patterns.In this research, LSTM-based framework with residual unit and attention mechanism is proposed to enhance the temporal dependencies and important features of stock price movements.Residual unit with skip connection captures more complex patterns and representations in stock price data and reduces the over-fitting problem to noisy time series data.LSTM with attention focuses on the significant time stamps which enhances the model prediction performance.The proposed system is experimented on five datasets: Apple (AAPL), Bitcoin, Ethereum, Litecoin and GOLD_PRICE.To prove the effectiveness of the model, the proposed system is compared with LSTM and Bidirectional LSTM (Bi-LSTM) models.Experimental results show that the proposed system outperforms baseline models such as LSTM, Bi-LSTM, LSTM+Bi-LSTM and state-of-the-art methods in term of error rates such as mean square error, root mean square error and mean absolute error.

Cătălina Cocianu, Cristian Răzvan Uscatu

Forecasting the price of cryptocurrencies is a notoriously hard and significant problem, due to the rapid market growth and high volatility. In this article, we propose a methodology for predicting future values of cryptocurrency exchange rates by developing a Non-linear Autoregressive with Exogenous Inputs (NARX) prediction model that uses the most adequate external information. The exogenous variables considered are historical values of the exchange rate and a series of technical indicators. The selection of the most relevant external inputs is based on the computation of the mutual information indicator and estimated using the k-nearest neighbor method. The methodology employs a fine-tuned Long Short-Term Memory (LSTM) neural network as the regressor. We have used quantitative and trend accuracy measures to compare the proposed method against other state-of-the-art LSTM-based models. In addition, regarding the input selection process, the proposed approach was compared against the most commonly used one, which is based on the cross-correlation coefficient. A long series of experiments and statistical analyses proved that the proposed methodology is highly accurate and the resulting model outperforms the state-of-the-art LSTM-based models.