This study assesses whether Bitcoin’s linkage with AI equities remains robust after accounting for equity risk sentiment. To this end, the study employs the multiscale quantile‐on‐quantile correlation (MSQQC) and multiscale quantile‐on‐quantile partial correlation (MSQQPC) approaches, using daily data covering 02/01/2019–16/06/2025. The results indicate that BTC–AI comovement is strongly state‐ and frequency‐dependent rather than stable across the joint distribution or across horizons. In the high‐frequency band, dependence is weak and only intermittently significant, with localised negative regions around BTC ≈ 0.20 with AI ≈ 0.30–0.50 and BTC ≈ 0.30 with AI ≈ 0.70. In the mid‐frequency band, significance concentrates in the tails, showing negative dependence under downside stress conditions such as BTC ≈ 0.10–0.30 with AI ≈ 0.10, alongside sign changes when BTC is in upper‐tail states. In the low‐frequency band, dependence becomes broadly positive and significant across most quantile combinations, with limited decoupling when AI is highly elevated (≈ 0.80–0.90) and BTC is also in upper quantiles (≈ 0.70–0.90). Importantly, conditioning on VIX and VVIX does not materially alter these patterns, suggesting that sentiment influences segments of short‐run dependence but does not overturn the longer‐run BTC–AI linkage. The study derives policy recommendations from these findings.
Το αυξανόμενο ενδιαφέρον για εναλλακτικές επενδύσεις που προσφέρουν διαφοροποίηση και υψηλότερες αποδόσεις, καθοδηγείται από τους περιορισμούς των παραδοσιακών χρηματοπιστωτικών αγορών και τις εξελισσόμενες ανάγκες των σύγχρονων επενδυτών. Στην παρούσα διδακτορική διατριβή διερευνώνται οι μηχανισμοί μεταβολής των τιμών δύο τομέων εναλλακτικών επενδύσεων: των επενδύσεων με κριτήρια Περιβάλλοντος, Κοινωνίας και Διακυβέρνησης (Environmental, Social, and Governance – ESG), με έμφαση στις εισηγμένες εταιρείες που επιδεικνύουν περιβαλλοντική υπευθυνότητα, και των ψηφιακών περιουσιακών στοιχείων που βασίζονται στην τεχνολογία Blockchain, μέσα από ένα ενοποιημένο μεθοδολογικό πλαίσιο. Το πρώτο μέρος της διατριβής εξετάζει τις χρηματοοικονομικές επιπτώσεις της Εταιρικής Περιβαλλοντικής Υπευθυνότητας (Corporate Environmental Responsibility – CER) στις επιχειρήσεις του δείκτη S&P 500 κατά τη διάρκεια δεκαπέντε ετών. Αξιολογείται η επίδραση της περιβαλλοντικής επίδοσης στην αποτίμηση της αγοράς μέσω δεικτών προσαρμοσμένων στον κίνδυνο, οι οποίοι βασίζονται στο υπόδειγμα CAPM και στην υπόθεση της αποτελεσματικής αγοράς. Η ανάλυση, η οποία στηρίζεται σε τεχνικές παλινδρόμησης δεδομένων πάνελ, όπως οι εκτιμήσεις OLS και 3SLS, εισάγει την έννοια του «Πράσινου Premium» — ενός μετρήσιμου αντισταθμίσματος μεταξύ περιβαλλοντικής υπευθυνότητας και αποδόσεων των επενδυτών. Τα αποτελέσματα δείχνουν ότι, ενώ η CER συσχετίζεται θετικά με λειτουργικούς δείκτες, όπως οι πωλήσεις και τα κέρδη, η χρηματιστηριακή απόδοση παραμένει κατώτερη, γεγονός που υποδηλώνει ένα διαρκές μειονέκτημα για τις επιχειρήσεις με περιβαλλοντικό προσανατολισμό στα μάτια των επενδυτών. Το δεύτερο μέρος της διατριβής επικεντρώνεται στη χρηματοοικονομική διάσταση του Blockchain, παρέχοντας μια ολοκληρωμένη ανάλυση της συμπεριφοράς τιμών των Μη Εναλλάξιμων Διακριτικών (Non-Fungible Tokens – NFTs) μέσω σύγχρονων μεθόδων μηχανικής μάθησης. Δεδομένα που αντλήθηκαν απευθείας από πλατφόρμες Blockchain και μετασχηματίστηκαν μέσω προηγμένων τεχνικών μηχανικής χαρακτηριστικών, χρησιμοποιούνται για την αξιολόγηση της προβλεπτικής ικανότητας των μοντέλων Random Forest, XGBoost και Πολυεπίπεδου Αντιληπτή (Multilayer Perceptron). Το τελικό σύνολο δεδομένων προέκυψε από μια προηγμένη διαδικασία μετασχηματισμού χαρακτηριστικών, εμπλουτισμένη με σύνθεση χαρακτηριστικών βάσει εξειδικευμένης γνώσης και Deep Feature Synthesis, αναδιαμορφωμένο μέσω Ανάλυσης Κύριων Συνιστωσών (Principal Component Analysis – PCA) και βελτιστοποιημένο μέσω τεχνικών επιλογής χαρακτηριστικών για βέλτιστη ερμηνευσιμότητα και μείωση διαστασιμότητας. Από τα μοντέλα που εξετάστηκαν, το XGBoost παρουσίασε τη μεγαλύτερη ακρίβεια πρόβλεψης, με τα ιστορικά δεδομένα τιμών να αποτελούν τον πιο καθοριστικό παράγοντα πρόβλεψης. Συνολικά, τα δύο μέρη της διατριβής συμβάλλουν στη διεύρυνση της βιβλιογραφίας σχετικά με τις μη παραδοσιακές κατηγορίες επενδυτικών στοιχείων, μέσω της εφαρμογής ισχυρών αναλυτικών μεθόδων σε διαφορετικά επενδυτικά πεδία. Τα αποτελέσματα παρέχουν εξειδικευμένες γνώσεις σχετικά με τους παράγοντες που διαμορφώνουν την αξία των περιουσιακών στοιχείων και θεμελιώνουν ένα μεταβιβάσιμο μεθοδολογικό πλαίσιο για την πρόβλεψη τιμών περιουσιακών στοιχείων σε αναδυόμενες και ετερογενείς αγορές.
Rachid Bourday, Ali Zaaouat, Issam Aattouchi, M. Ait Kerroum
Predicting cryptocurrency prices with precision is crucial for strategic financial planning, enabling stakeholders to mitigate risks in the highly unpredictable nature of digital assets. This paper presents an innovative framework combining Bidirectional Long Short-Term Memory (Bi-LSTM) and Graph Attention Networks (GATs) to improve forecasting accuracy for Ethereum. The Bi-LSTM analyzes time-based trends in historical price and trading volume over a 90-day horizon, whereas GATs examine correlations between critical market features, including 20-day and 30-day moving averages, through attention-focused techniques. When applied to Ethereum's historical price data, the model achieves an MSE of 0.0021, RMSE of 0.046, and MAE of 0.032, exceeding traditional LSTM-based approaches. These outcomes highlight the advantages of fusing sequential neural architectures with graph-structured relational modeling to refine predictive accuracy. By unifying time-series and graph-structured data analysis, this study contributes to the advancement of financial analytics powered by deep learning, equipping traders and researchers with an actionable tool to refine trading tactics within Ethereum's dynamic ecosystem.
This paper examines whether social media sentiment derived from Twitter and Reddit improves the explanation and prediction of cryptocurrency volatility. Using Bitcoin and Ethereum as benchmark assets, we combine sentiment indicators with GARCH-type models and the HAR-RV framework. Results suggest that cryptocurrency volatility is primarily driven by internal market dynamics rather than social media sentiment.
This paper investigates the extent to which cognitive heuristics, social influence, and digitally-mediated sentiment drive the extreme volatility of cryptocurrency, Decentralised Finance (DeFi), and Non-Fungible Token (NFT) markets, and the degree to which these dynamics deviate from the Efficient Market Hypothesis. Using an integrative narrative review and a synthesis of empirical evidence from 2014–2025, the paper develops the Integrated Digital Asset Behavioural Model (IDABM), a four-variable framework relating market stability to social velocity (Sv ), heuristic load (Hl ), platform gamma (Pγ ), and liquidity leverage (Ll ). The analysis draws on demographic and sentiment data, case evidence from the 2022 Terra/ Luna and FTX collapses, and a comparative cross-asset bias taxonomy. The findings indicate that digital asset markets constitute a pure sentiment environment in which the absence of conventional valuation anchors produces heuristic dominance and structurally amplified herding behaviour. The paper concludes that effective regulation must shift from informational disclosure toward behavioural guardrails — including algorithmic accountability, regulation of gamified trading interfaces, and behavioural literacy requirements.
Digitalisation of finance led to the creation of a digital financial economy, where digital assets such as cryptocurrencies, decentralized financial assets, non-fungible tokens, stablecoins, etc. were traded. In this study, machine learning and deep learning techniques, including ARIMA, FB Prophet, LSTM, and BiLSTM, have been used to forecast the prices of digital assets. In this study, Bitcoin, Ethereum, Uniswap, Aave, ApeCoin, and Decentraland tokens have been categorized into three groups, and the prediction models have been trained using the tokens' closing prices. The authors find that NFTs have been underestimated and that DeFi assets have greater growth potential. Whereas cryptocurrencies have been traded more and shown greater volatility than other asset classes. BiLSTM achieves the best results, with higher accuracy in price prediction. Here, it has been seen that ApeCoin, Decentraland, and Bitcoin are more stable than other assets. Thus, for an optimised portfolio and additional savings, it is necessary to provide a proper asset mix.
We document stable cross-asset patterns in cryptocurrency limit-order-book microstructure: the same engineered order book and trade features exhibit remarkably similar predictive importance and SHAP dependence shapes across assets spanning an order of magnitude in market capitalization (BTC, LTC, ETC, ENJ, ROSE). The data covers Binance Futures perpetual contract order books and trades on 1-second frequency starting from January 1st, 2022 up to October 12th, 2025. Using a unified CatBoost modeling pipeline with a direction-aware GMADL objective and time-series cross validation, we show that feature rankings and partial effects are stable across assets despite heterogeneous liquidity and volatility. We connect these SHAP structures to microstructure theory (order flow imbalance, spread, and adverse selection) and validate tradability via a conservative top-of-book taker backtest as well as fixed depth maker backtest. Our primary novelty is a robustness analysis of a major flash crash, where the divergent performance of our taker and maker strategies empirically validates classic microstructure theories of adverse selection and highlights the systemic risks of algorithmic trading. Our results suggest a portable microstructure representation of short-horizon returns and motivate universal feature libraries for crypto markets.
This study introduces a novel approach to walk-forward optimization by parameterizing the lengths of training and testing windows. We demonstrate that the performance of a trading strategy using the Exponential Moving Average (EMA) evaluated within a walk-forward procedure based on the Robust Sharpe Ratio is highly dependent on the chosen window size. We investigated the strategy on intraday Bitcoin data at six frequencies (1 minute to 60 minutes) using 81 combinations of walk-forward window lengths (1 day to 28 days) over a 19-month training period. The two best-performing parameter sets from the training data were applied to a 21-month out-of-sample testing period to ensure data independence. The strategy was only executed once during the testing period. To further validate the framework, strategy parameters estimated on Bitcoin were applied to Binance Coin and Ethereum. Our results suggest the robustness of our custom approach. In the training period for Bitcoin, all combinations of walk-forward windows outperformed a Buy-and-Hold strategy. During the testing period, the strategy performed similarly to Buy-and-Hold but with lower drawdown and a higher Information Ratio. Similar results were observed for Binance Coin and Ethereum. The real strength was demonstrated when a portfolio combining Buy-and-Hold with our strategies outperformed all individual strategies and Buy-and-Hold alone, achieving the highest overall performance and a 50 percent reduction in drawdown. A conservative fee of 0.1 percent per transaction was included in all calculations. A cost sensitivity analysis was performed as a sanity check, revealing that the strategy's break-even point was around 0.4 percent per transaction. This research highlights the importance of optimizing walk-forward window lengths and emphasizing the value of single-time out-of-sample testing for reliable strategy evaluation.
The book Solutions and Technologies for Modern Business stands as a relevant contribution to understanding the technological and strategic transformations impacting the contemporary business environment.With a broad approach, the work offers reflections on innovative solutions, technological tools, and management practices aimed at strengthening and adapting organizations in the face of constant market changes.By emphasizing the integration of theory and practice, the book contributes to the development of critical analyses regarding the use of technology in organizational processes, highlighting its importance for competitiveness, innovation, and decision-making.The work brings together diverse perspectives that enrich academic debate and encourage the development of more efficient and sustainable strategies in the business context.This book is recommended for professors, students, and professionals in the fields of administration, management, technology, and business, as well as for anyone interested in expanding their knowledge of solutions and technologies applied to the corporate environment.It is a work that fosters learning, reflection, and the improvement of organizational practices in the digital era.
Bitcoin’s price dynamics are influenced by both internal factors (e.g., supply shocks, investor sentiment) and external drivers, among which the stability of stablecoins has attracted increasing academic and regulatory attention. This paper investigates the effect of stablecoin peg deviations (USDT and USDC) on Bitcoin returns using daily data from January 2020 to August 2025. Based on a vector autoregression (VAR) framework, we conduct unit root tests, lag order selection, model estimation, Granger causality tests, and impulse response analysis. Results show that both Bitcoin returns and stablecoin deviations exhibit strong short-term inertia. USDT and USDC deviations significantly Granger-cause Bitcoin returns, whereas the reverse causality is weaker. Impulse responses indicate that stablecoin deviations first produce positive shocks to Bitcoin returns, followed by negative corrections that gradually stabilize. The effect of USDT is more pronounced and persistent, underscoring its central role in cryptocurrency markets. These findings highlight the importance of monitoring stablecoin market stability, especially USDT, for investors and regulators seeking to manage systemic risks in crypto markets.
This thesis investigates the design of automated market makers (AMMs) for trading tokenized derivatives in decentralized finance. Motivated by the limitations of existing AMMs, which only permit strictly positive prices, we introduce an invariant that also allows for negative prices. As a primary use case, the thesis develops an AMM for trading an offset token against tokenized euros. The thesis employs, on the one hand, a Monte Carlo–based simulation to evaluate the risk-adjusted returns of an AMM. The simulation includes two types of traders: arbitrage traders, who exploit price deviations between the AMM and the fair value, and noise traders, who represent demand for liquidity. On the other hand, we introduce KPIs such as impermanent loss and market depth. The goal of the thesis is to analyse whether these KPIs can be used to predict the risk-adjusted returns of an AMM.
Cryptocurrencies have upended the financial industry since they provide decentralized and peer-to-peer transactions. However, due to market volatility and the numerous non-linear relationships between price dynamics and human mood, forecasting Bitcoin values is a difficult task. The deep learning architecture shown in this work combines sentiment confidence scores derived from cryptocurrency-related tweets utilizing Transformer-based natural language processing with historical price indicators. The model incorporates Convolutional Neural Networks (CNN) to detect local time-series patterns and Long Short-Term Memory (LSTM) networks to produce long-term dependencies. We apply this architecture, involving sequence-based preprocessing and normalization, to Bitcoin and Ethereum to ensure robustness. Evaluations in comparison to baseline models Sentiment fusion dramatically increases predicting accuracy, especially during times of market turbulence, according to CNN-LSTM without sentiment, vanilla LSTM, and ARIMA. Our research helps develop scalable, sentiment-aware financial forecasting algorithms that better reflect the behavior of real markets.
This study systematically reviews scientific research on predicting cryptocurrency markets. A total of 790 articles obtained from the Web of Science database were included in the analysis, and the structure of the literature was evaluated using bibliometric methods. The preliminary investigation indicated that studies examining the prediction of cryptocurrencies have undergone a substantial increase since 2016. While a significant proportion of the extant literature pertains to Bitcoin, the first cryptocurrency, it is evident that other cryptocurrencies, such as Ethereum, have also attracted the attention of researchers over the years. The analysis yielded four primary categories: machine learning-based prediction methods, financial risk and volatility analyses, behavioral and technical determinants, and finally, advanced deep learning methods. In the context of cryptocurrency prediction, studies have underscored the significance of attributes, emphasizing their role in enhancing the efficacy of prediction models. These studies have also highlighted the impact of integrating machine learning and deep learning-based models with conventional methods in enhancing the performance of established models. The study emphasizes the necessity to direct future research towards the integration of behavioral indicators and the examination of multiple market relationships.
This study aims to analyze the volatility spillovers between Bitcoin and Ethereum, the two main actors in the cryptocurrency market, and altcoins across sectoral and financial groups. Using data from January 1, 2021, to March 6, 2023, the study applied the VAR-based method developed by Diebold and Yılmaz (2012) and measured both directional and total volatility spillovers. The findings show that Bitcoin's volatility largely stems from internal dynamics and spreads to other cryptocurrencies to a limited extent. In contrast, Ethereum is more affected by external shocks and exhibits a stronger volatility spillover across the market. Among altcoin categories, Gaming, Analytics, and DeFi groups were found to be the most influential in volatility transmission, while thematic tokens such as NFT, Web3, and Metaverse were more sensitive to external volatility. In contrast, stablecoins and tokens in the identity and healthcare sectors were found to have relatively low volatility and a more stable structure. These results offer important insights for investors and regulators regarding risk management strategies and portfolio diversification. The study provides a valuable framework for understanding the systematic volatility dynamics within the cryptocurrency ecosystem
This research investigates the predictive power of news sentiment from Google News on Bitcoin price movements, leveraging a five-year dataset of news headlines (2019 to 2024). By correlating sentiment scores with historical Bitcoin prices, the study employs various machine learning algorithms to forecast price trends. The results indicate that while Decision Tree and Random Forest models offer balanced predictions, Logistic Regression and Support Vector Machines achieve high AUC scores but suffer from class imbalance. In contrast, Naïve Bayes and KNN models prove less effective. The findings suggest that sentiment analysis of news headlines can provide moderate short-term predictions for Bitcoin price fluctuations. This study introduces an innovative tool for investors and market analysts, offering insights into the influence of news sentiment on cryptocurrency prices.
Deep learning has emerged as a widely applied approach across various fields, with finance and forecasting being among its most prominent areas of use. Within this domain, different deep learning architectures have been developed to address specific prediction problems. This study compares the performance of ARIMAX and several deep learning models—including LSTM, BILSTM, CNN-LSTM, GRU, and TFT—in forecasting Bitcoin prices. The dataset consists of daily values from January 2014 to January 2025. The dependent variable is the daily Bitcoin closing price ($), while the independent variables include oil price (USD/barrel), gold price (USD/ounce), platinum price ($/XPT), and the USD/TRY exchange rate. All analyses were conducted in Python using Google Colab, with the Keras library employed for model implementation. Root Mean Square Error (RMSE) was selected as the evaluation metric for predictive accuracy. The results indicate that the TFT model achieved the highest predictive performance, followed closely by the GRU model. LSTM, BILSTM, and ARIMAX models showed similar yet weaker performance, while the CNN-LSTM model produced the least accurate forecasts, with significantly higher RMSE values compared to the other models.
Virtual currency has become one of the most sought-after alternative assets in the past decade with bitcoin being a leading example. value leapt from its starting price of $0.0025 to increase by more than 40 million times that amount, creating one of the greatest rises in value in the entire history of finance. In the past few years, many academic studies show that even though Bitcoin runs independently from traditional finance, but still there is a high correlation between Bitcoin and stock market. In particular, following the introduction of Bitcoin options back in 2017, Bitcoin now appears more predictive of stock return movements than before. Research by Afees A. Salisu and his coworkers display that a solitary Bitcoin price prediction model using an optimized predictive regression framework notably surpasses older ones. but don’t say how long this goes on Therefore this research will go to try and determine the time frame when Bitcoin is better at predicting the future of the stock market as opposed to stock options. Also, we’ll use machine learning techniques to train machine learning models to predict the movements of the stock market and see if they work.
This study introduces a unified and methodologically symmetric comparative framework for multivariate cryptocurrency forecasting, addressing long-standing inconsistencies in prior research where model families, feature sets, and preprocessing pipelines differ across studies. Under an identical and rigorously controlled experimental setup, we benchmark six deep learning architectures—LSTM, GPT-2, Informer, Autoformer, Temporal Fusion Transformer (TFT), and a Vanilla Transformer—together with four widely used econometric models (ARIMA, VAR, GARCH, and a Random Walk baseline). All models are evaluated using a shared multivariate feature space composed of more than forty technical indicators, identical normalization procedures, harmonized sliding-window formations, and aligned temporal splits across five high-liquidity assets (BTC, ETH, XRP, XLM, and SOL). The experimental results show that transformer-based architectures consistently outperform both the recurrent baseline and classical econometric models across all assets. This superiority arises from the ability of attention mechanisms to capture long-range temporal dependencies and adaptively weight informative time steps, whereas recurrent models suffer from vanishing-gradient limitations and restricted effective memory. The best-performing deep learning models achieve MAPE values of 0.0289 (BTC, GPT-2), 0.0198 (ETH, Autoformer), 0.0418 (XRP, Informer), 0.0469 (XLM, Informer), and 0.0578 (SOL, TFT), substantially improving upon the performance of both LSTM and all econometric baselines. These findings highlight the effectiveness of attention-based architectures in modeling volatility-driven nonlinear dynamics and establish a reproducible, symmetry-preserving benchmark for future research in deep-learning-based financial forecasting.
ABSTRACT Based on the rationale that returns and volatility are interrelated, we apply a multilayer network framework involving the return layer and volatility layer of cryptocurrencies, NFTs, and DeFi assets over the period January 1, 2018–January 23, 2024. The results show significant connectedness in each of the return and volatility layers, with major cryptocurrencies such as Bitcoin and Ethereum playing a central role. Large spikes in the level of connectedness are noticed around COVID‐19 pandemic and Russia–Ukraine conflict, and Bitcoin and Ethereum emerge as net transmitters of returns and volatility shocks, emphasizing their significant role around these crisis periods. Notably, a strong positive rank correlation exists between the return and volatility layers, highlighting the significant risk–return relationship in the digital asset class. The findings suggest that economic actors should not ignore the interconnectedness between the return and volatility layers in the system of cryptocurrencies, NFTs, and DeFi assets for the sake of a comprehensive analysis of information flow. Otherwise, a share of the information flow concerning the return–volatility nexus across these digital assets would be missed, possibly leading to inferences regarding asset pricing, portfolio allocation, and risk management.