This study proposes a framework combining Variational Mode Decomposition (VMD) with a relevance-driven selection process to reduce noise and redundancy in financial time-series forecasting. The original time series is decomposed by VMD into intrinsic mode functions (IMFs), which are then evaluated using three relevance metrics: relative energy contribution, mutual information, and Spearman's rank correlation coefficient. These metrics identify the IMFs most strongly associated with future price movements. As opposed to conventional VMD-based approaches that treat all IMFs equally, the proposed relevance-driven selection process adapts IMF selection to the statistical properties of the analyzed market, thereby improving model generalization across different volatility conditions and forecasting horizons. This study makes three main contributions: (i) developing a relevance-driven IMF selection strategy to overcome limitations of traditional VMD methods, (ii) designing a hybrid framework that integrates multiscale decomposition with nonlinear information filtering, and (iii) conducting a comprehensive empirical evaluation of the proposed models. Experiments on hourly Bitcoin (BTC)/USD data from 2018 to 2025 show that the VMD-RDIC-deep learning models achieves strong forecasting performance. The results show that the proposed relevance-driven decomposition framework improves prediction accuracy and robustness compared with traditional statistical models, including Autoregressive Integrated Moving Average (ARIMA), as well as machine learning and deep learning approaches, highlighting its suitability for complex and volatile financial markets. Received: 18 January 2026 | Revised: 13 April 2026 | Accepted: 23 July 2026 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement The data that support the findings of this study are openly available at https://www.kaggle.com/datasets/novandraanugrah/bitcoin-historical-datasets-2018-2024. Author Contribution Statement Maryam Maatallah: Conceptualization, Methodology, Software, Data curation, Writing – original draft, Visualization. Mourad Fariss: Software, Formal analysis, Writing – original draft. Hakima Asaidi: Validation, Investigation, Writing – review & editing. Mohamed Bellouki: Resources, Writing – review & editing, Supervision, Project administration.
Dr. Mohamed Nachat, Hassan Oukhouya, Saïd El Melhaoui, Moustapha Faizi · 7 authors
Time series forecasting plays a central role in finance, energy, and public health. Classical statistical, machine learning, deep learning, and generative approaches have all been applied to forecasting tasks in these fields, but comparisons between them are usually confined to a single domain or to models from the same family, and few studies report both univariate and multivariate results under the same conditions. This paper presents a controlled cross-domain comparison of four representative paradigms: classical statistics (Seasonal Autoregressive Integrated Moving Average with Exogenous variables, SARIMAX), gradient boosting machine learning (Light Gradient Boosting Machine, LightGBM), recurrent deep learning (Recurrent Neural Network, RNN), and generative-adversarial deep learning (Conditional Generative Adversarial Network, CGAN). Each model is evaluated on three monthly datasets with contrasting characteristics: Bitcoin prices (175 observations, high volatility), U.S. energy consumption (612 observations, strong seasonality), and U.S. cardiovascular mortality (300 observations, gradual trend with pandemic shock). Both univariate and multivariate variants are tested under the same preprocessing and one-step-ahead evaluation protocols, using eight performance metrics. The CGAN reaches the lowest MAPE on energy consumption (2.88%). On Bitcoin, the multivariate LightGBM lowers the MAPE from 28.26 to 19.25%, while on cardiovascular mortality the RNN reaches 3.34% MAPE. No paradigm performs best in every domain, and the gain from exogenous variables depends on both the paradigm and the domain.
Blockchain users pay transaction fees to miners or block proposers who validate and add transactions to the distributed ledger. Ethereum introduces the concept of gas to decouple transaction costs from Ether's price volatility, calculating fees based on gas units. The current mechanism defined by Ethereum Improvement Proposal (EIP) 1559 dynamically adjusts the base fee according to block gas usage. However, its rule-based adjustment can lead to unstable gas consumption when demand fluctuates within a narrow range and struggles to respond efficiently to sudden demand spikes, such as during non-fungible token (NFT) drops. To address these limitations, we propose a deep reinforcement learning-based transaction fee mechanism that learns an adaptive base-fee update policy. Our approach maintains gas consumption close to the target level across various demand scenarios and stabilizes transaction fees and gas usage per block even under abrupt demand shifts. These results demonstrate that the proposed method provides a more adaptive and resilient fee adjustment mechanism compared to the current EIP-1559 model.
The housing market is of great significance to the development and advancement of cities, but customary forms of property valuation are frequently biased, time-consuming, and not always effective. This paper focuses on the city of Irbid in Jordan, aiming to collect all the information on apartments and houses, predict the prices of properties, and clarify the key factors influencing the prices. Following the comprehensive cleaning process of the data and exploratory analysis, three ensemble machine learning models were trained and optimized to achieve accurate price predictions. The performance of all three models demonstrated excellent and consistent predictions, highlighting the efficiency of ensemble methods in predicting property prices. SHAP analysis indicated that the size of the house, the number of bedrooms, the number of lounges as well as the location are the most significant factors influencing the prices in Irbid. This reflects the functioning of the local market.
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 ANN–LSTM 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.
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 Bitcoin’s 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 Bitcoin’s 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 Bitcoin’s price, along with insights into the factors affecting its price movements. The proposed LSTM model has achieved 98% accuracy.
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.
The integration of renewable energy sources (RES) into modern power grids has enabled decentralized energy generation at the community level, fostering peer-to-peer (P2P) energy trading among prosumers and microgrids. Accurate forecasting of household energy consumption and photovoltaic (PV) generation is critical for optimizing energy flows, enhancing grid reliability, and enabling cost-effective trading decisions. This paper presents an intelligent energy trading platform that integrates machine learning-based forecasting, battery-aware decision-making, and blockchain-enabled transactions to facilitate secure and efficient local energy exchange. Using historical smart meter and weather data from London households, multiple forecasting models including GRU, LSTM, Random Forest, and XGBoost were trained and evaluated. The GRU model achieved superior performance in predicting energy consumption, while Random Forest produced the most accurate PV generation forecasts. These predictions were combined with household battery levels to dynamically determine next-day operational roles: Buyer, Seller, Store, or Use Battery. Unlike conventional fixed-threshold approaches, the framework supports user-defined variable battery thresholds, allowing personalized energy management strategies. The proposed decision-making model achieved an accuracy of 90.72 % for one random block, and extended simulations across 29 different random household blocks confirmed its robustness with an average accuracy of 88.69 % (95 % CI: 87.9–89.6 %). In the trading phase, households participate in a decentralized energy trading platform powered by blockchain and smart contracts. Based on the next-day forecasts, a linear programming-based optimization algorithm matches buyer requests and seller offers to minimize the total system cost while ensuring fairness and efficient energy allocation. To assess its performance, the proposed optimization approach was compared against a greedy matching algorithm where sequential matching is done without a cost optimization and a grid baseline scenario where no storage/sharing of energy takes place. The optimized matching consistently achieved substantially lower trading costs across all households demonstrating superior efficiency, fairness, and scalability compared to the benchmark methods. All transactions are executed securely and transparently on the blockchain through Ethereum-based smart contracts, which automate energy trading, pricing, and settlement. A user-friendly web interface was developed to allow participants to monitor and interact seamlessly with the platform. Overall, this battery-aware, community-driven trading framework showcases how intelligent energy forecasting, cost-optimized decision-making, and blockchain-enabled trading can collectively enhance energy autonomy, cost savings, and renewable energy utilization at both the household and community levels.
The increasing complexity of urban energy systems requires decentralized, sustainable, and scalable solutions. The paper presents a new multi-layered framework for smart energy management in microgrids by bringing together advanced forecasting, decentralized decision-making, evolutionary optimization and blockchain-based coordination. Unlike previous research addressing these components separately, the proposed architecture combines five interdependent layers that include forecasting, decision-making, optimization, sustainability modeling, and blockchain implementation. A key innovation is the use of Temporal Fusion Transformer (TFT) for interpretable multi-horizon forecasting of energy demand, renewable generation, and electric vehicle (EV) availability which outperforms conventional LSTM, GRU and RNN models. Another novelty is the hybridization of Genetic Algorithms (GA) and Particle Swarm Optimization (PSO), to simultaneously support discrete and continuous decision variables, allowing for dynamic pricing, efficient energy dispatching and adaptive EV scheduling. Multi-Agent Reinforcement Learning (MARL) which is improved by sustainability shaping by including carbon intensity, renewable utilization ratio, peak to average load ratio and net present value in agent rewards. Finally, Ethereum-based smart contracts add another unique contribution by providing the implementation of transparent and tamper-proof peer-to-peer energy trading and automated sustainability incentives. The proposed framework strengthens resilient infrastructure through decentralized coordination and intelligent optimization while contributing to climate mitigation by reducing carbon intensity and enhancing renewable integration. Experimental results demonstrate that the proposed framework achieves a 14.6% reduction in carbon intensity, a 12.3% increase in renewable utilization ratio, and a 9.7% improvement in peak-to-average load ratio compared with baseline models. The TFT-based forecasting model achieves RMSE = 0.041 kWh and MAE = 0.032 kWh, outperforming LSTM and GRU by 11% and 8%, respectively.
The increasing integration of renewable energy into smart grids introduces challenges of demand-supply imbalance, peak load stress, and cyber-physical vulnerabilities. Existing demand response (DR) frameworks often lack scalability, privacy-preserving data sharing, and secure transaction mechanisms, which limit user participation and grid resilience. To address these challenges, this study proposes GridSyncNet, a blockchain-enabled multi-agent deep reinforcement learning framework for real-time demand response. The framework integrates federated learning to enhance decentralized forecasting accuracy, blockchain consensus to ensure transparent and tamper-proof energy trading, and actor–critic based DRL agents to dynamically optimize load scheduling and energy dispatch across prosumers. Extensive simulations demonstrate that GridSyncNet outperforms benchmark models such as OD-CNN, D-FCAS, and USTCF. Specifically, it achieves a 98.2 % demand response efficiency, 30.6 % reduction in carbon emissions, and 97.4 % forecasting accuracy. Comparative analysis with multi-agent DRL (MADRL) approaches further confirms that GridSyncNet provides superior scalability, privacy, and security in decentralized environments. The proposed framework contributes to the design of secure, resilient, and sustainable energy management systems, offering practical insights for accelerating the transition toward net-zero energy communities. By combining blockchain, federated learning, and multi-agent reinforcement learning, GridSyncNet establishes a comprehensive pathway for trustworthy and adaptive smart grid operations. • A multi-agent deep reinforcement learning framework for adaptive DR in smart grids. • Decentralized peer-to-peer energy trading to transparent, secure energy trading. • Renewable energy utilization 89 %, CO 2 reduction 30.6 % & forecasting accuracy 92.4 %. • Federated learning & improved system resilience against cyber threats for DSM. • Optimize load balancing, peak shaving & cost efficiency for distributed grid agents.
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 Naïve methods.
Investment advisory services are now commonly offered by consulting firms with financial experts, typically for a monthly fee. Financial markets require specialized knowledge, but advancements in artificial intelligence have revolutionized this field. Deep learning algorithms, especially Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU), are widely used to predict asset price trends in nonlinear time-series data. However, they demand large datasets and are prone to overfitting. Recently, combining deep learning with reinforcement learning has shown promise, though it requires intensive research and computational resources. This study introduces the BTC-PDPR (Bitcoin Price Direction Prediction Robot) model, which predicts Bitcoin's daily price direction using the Random Forest Regressor. As an ensemble-based machine learning model, it works effectively with smaller datasets and identifies key technical indicators influencing price trends. The model achieved a 99.20% accuracy rate on data from March 2018 to the present. It runs efficiently in Google Colab (v5e1 configuration), producing results in just 22 seconds. This paper outlines the methodology, reviews relevant studies from 2017 to 2024, highlights gaps in the literature, and emphasizes the study’s contributions to the field.
Bitcoin's high volatility poses significant challenges for short-term price prediction, making it a critical area of study for financial forecasting. Traditional models such as Long Short-Term Memory (LSTM) networks often encounter difficulties in handling long-range dependencies and non-stationary data, limiting their predictive accuracy under volatile conditions. This study introduces the Time-Series Transformer (TST) as a novel approach to predict Bitcoin's short-term prices. By leveraging self-attention mechanisms, TST effectively captures complex temporal patterns in historical Bitcoin data, including prices and trading volume. The data was segmented into fixed-length windows to facilitate model training and testing. Evaluation metrics such as Mean Squared Error (MSE), Mean Absolute Scaled Error (MASE), and R-squared (R²) demonstrated TST’s superior performance over LSTM, particularly during periods of high market fluctuation. Furthermore, TST exhibited notable computational efficiency when working with large datasets, underscoring its scalability. These findings not only highlight TST’s potential for enhancing cryptocurrency price prediction but also pave the way for future research integrating external data sources and exploring further model enhancements for more robust financial forecasting.
The stock market is a vital component of the financial sector. Due to the inherent uncertainty and volatility of the stock market, stock price prediction has always been both intriguing and challenging. To improve the accuracy of stock predictions, we construct a model that integrates investor sentiment with Long Short-Term Memory (LSTM) networks. By extracting sentiment data from the “Financial Post” and quantifying it with the Vader sentiment lexicon, we add a sentiment index to improve stock price forecasting. We combine sentiment factors with traditional trading indicators, making predictions more accurate. Furthermore, we deploy our system on the blockchain to enhance data security, reduce the risk of malicious attacks, and improve system robustness. This integration of sentiment analysis and blockchain offers a novel approach to stock market predictions, providing secure and reliable decision support for investors and financial institutions. We deploy our system and demonstrate that our system is both efficient and practical. For 312 bytes of stock data, we achieve a latency of 434.42 ms with one node and 565.69 ms with five nodes. For 1700 bytes of sentiment data, we achieve a latency of 1405.25 ms with one node and 1750.25 ms with five nodes.
Tarihin başlangıcından itibaren sürekli evrim geçiren para, insanlık tarafından geliştirilen en önemli araçlardan biridir. Para, insanların gelecekteki ve anlık ihtiyaçlarını karşılamak için belirlenen bir değeri temsil eder. Para kavramı, dönemin koşullarına ve imkânlarına göre farklı şekillerde ortaya çıkar. Kripto paraların temelleri 1980'lerde atılmış olup, 2008 yılında Satoshi Nakamoto tarafından Bitcoin'in tanıtılmasıyla hayatımıza girmiştir. Geleneksel paralara alternatif olarak ortaya çıkan kripto paralar, teknolojik bir yenilik olup her geçen gün daha da popüler hale gelmektedir. Bitcoin, merkezi bir otorite tarafından yönetilmeyen ilk kripto paradır ve popülerliği ve başarısı diğer alternatif kripto paraların oluşmasına yol açmıştır. Julong Deng tarafından 1982 yılında geliştirilen “Gri Sistem Teorisi”, belirsiz sistemlerin davranışlarını tahmin etmek için kullanılan bir yöntem olup GM (1,1) modeli en sıklıkla kullanılan gri modeldir. Bu çalışma, Bitcoin'in fiyatlarını GM (1,1) modeli kullanarak tahmin etmeyi amaçlamaktadır. Araştırma sonucunda, modelin gelecek dönem tahminleri için uygun olduğu ve başarılı tahminler yaptığı belirlenmiştir
Moch Panji Agung Saputra, Riza Andrian Ibrahim, Renda Sandi Saputra
The increasing use of cryptocurrencies has changed the dynamics of investment, presenting both opportunities and challenges for investors. Although various studies have compared the performance of Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) in predicting financial asset prices, there are still differences in results regarding which model is superior. Therefore, this study aims to compare the performance of LSTM and GRU in predicting Ethereum prices using a hyperparameter tuning approach. The data used is historical data of Ethereum (ETH) shares from 2020 to 2025. The research methodology includes data preprocessing using Min-Max scaling, model development with various layer configurations, and comprehensive evaluation using several performance metrics. The results show that the GRU Model provides superior performance with a lower Root Mean Squared Error (RMSE) of 0.0234 and Mean Absolute Error (MAE) of 0.0168, compared to LSTM's RMSE of 0.0265 and MAE of 0.0193. While LSTM exhibits a slightly better Mean Absolute Percentage Error (MAPE) of 18.08% compared to GRU at 18.17%, the GRU model achieves a higher R² Score of 0.9442 compared to LSTM at 0.9282. Visual analysis of the prediction patterns and residual distributions further demonstrates GRU’s more consistent and accurate performance in capturing Ethereum price movements. These findings suggest that while both models are effective for cryptocurrency price prediction, GRU offers slightly better overall performance and stability, especially in maintaining consistent prediction accuracy across different market conditions.
Anak Agung Surya Pradhana, Kadek Suarjuna Batubulan
Forecasting Ethereum price changes presents challenges due to the cryptocurrency market’s volatility and rapid fluctuations. This study applies Long Short-Term Memory (LSTM) networks to predict Ethereum price trends using hourly historical data. The LSTM model captures temporal dependencies effectively, achieving moderate accuracy with a Root Mean Squared Error (RMSE) of 11.42. It performs well in stable market conditions, with predicted prices closely aligning with actual values, validating its potential for identifying long-term trends. However, the model struggles during high-volatility periods, failing to predict abrupt price spikes and market crashes accurately. Overfitting is also observed, indicated by disparities between training and test errors, limiting the model’s generalizability to unseen data. To address these issues, this research suggests incorporating features such as trading volumes, market sentiment, macroeconomic indicators, and blockchain metrics to enhance predictive accuracy. Additionally, employing advanced architectures like attention mechanisms, hybrid models, and real-time learning frameworks is recommended to improve adaptability and robustness in dynamic market environments. These enhancements aim to create a more comprehensive and reliable predictive tool. This study contributes to the advancement of predictive analytics in cryptocurrency markets, offering valuable insights for traders, investors, and policymakers navigating the complexities of digital finance.
The inherent challenges of financial time series forecasting demand advanced modeling techniques for reliable predictions. Effective financial time series forecasting is crucial for financial risk management and the formulation of investment decisions. The accurate prediction of stock prices is a subject of study in the domains of investing and national policy. This problem appears to be challenging due to the presence of multi-noise, nonlinearity, volatility, and the chaotic nature of stocks. This paper proposes a novel financial time series forecasting model based on the deep learning ensemble model LSTM-mTrans-MLP, which integrates the long short-term memory (LSTM) network, a modified Transformer network, and a multilayered perception (MLP). By integrating LSTM, the modified Transformer, and the MLP, the suggested model demonstrates exceptional performance in terms of forecasting capabilities, robustness, and enhanced sensitivity. Extensive experiments are conducted on multiple financial datasets, such as Bitcoin, the Shanghai Composite Index, China Unicom, CSI 300, Google, and the Amazon Stock Market. The experimental results verify the effectiveness and robustness of the proposed LSTM-mTrans-MLP network model compared with the benchmark and SOTA models, providing important inferences for investors and decision-makers.
In the past three decades, there has been a sweeping trend in Western and developed countries worldwide to transform the vertically integrated electricity supply chain into competitive electricity markets to diversify investment in the system and ultimately drive down operation costs. Nonetheless, due to some geopolitical and economic reasons, many developing countries adopted a modestly liberalized version of the power market (imperfect market). With the trend of privatization, specifically at the generation level, to leverage the hypothetical competitiveness, countries that did not adopt a full-fledged market structure face a dilemma. The system operators of incumbent imperfect market models find it increasingly difficult to deal with multiple private ownership of Independent Power Producers who are unwilling to share their detailed operational parameters for long-term generation scheduling (lasting for years). In this paper, Blockchain (BC) is being advocated as a platform that simulates a virtual market environment to address such issues. The proposed BC-based structure allows generators to participate in the short-term scheduling mechanism (such as day-ahead) in a trust-free environment without sharing their vital data yet achieving efficient, market-grade solutions. The feasibility of this new proposition is demonstrated through three different application scenarios, utilizing real-world load and renewable generation profiles sourced from the respective Grid System Operators databases. Python library (PYPSA) and Ethereum Testnet are being used for grid simulation and BC platform implementation respectively. The results of BC-assisted generation scheduling are presented and compared with the imperfect market model to highlight the viability of the proposed new approach.
The precise forecast of cryptocurrency prices is essential for portfolio investment because of their volatility and operability in virtual trading markets, which poses a huge trouble to investors' decision-making ability and investment planning. In this paper, we concentrated on the prediction of the recent trends of mainstream cryptocurrencies and selected them to optimize the portfolio to maximize profit and reduce risk. We used neural networks to solve this problem, which has three layers, including the LSTM layer, dropout layer, and dense layer. We focused on BCH, BTC, ETH, ETC, LTC, EOS, and XRP and collected their datasets for estimations. An LSTM model, a multi-task learning model, and a novel loss function, where the Negative Sharpe Ratio is provided, were implemented to predict the best portfolio (weights) for the cryptocurrencies mentioned above. The common evaluation indices, such as MSE, RMSE, MAE, and R-square ($ {R}^{2} $), can demonstrate the accuracy and reliability of the Neural Network models. Due to the significant price differences among currencies, the values of MSE, MSE, and MAE were large, making it difficult to evaluate their accuracy. Therefore, $ {R}^{2} $ was adopted. Finally, we simulated the portfolio investment and saw the revenue compared to the existing approaches. The new machine learning model abandoned the previous methods, which only predicted and analyzed a single cryptocurrency and ignored the correlations among them. Therefore, our innovation of this research was to use neural networks to consider investment plans combining multiple currencies while minimizing volatility and ensuring a large Sharpe Ratio, thereby obtaining the best portfolio investment of cryptocurrencies.
We propose a GA-optimized self-attention LSTM (SAG-LSTM) for multi-asset price forecasting and evaluate it on daily series of crude oil, gold, and Bitcoin, augmented with trading volumes (01-Apr-2021 to 30-Dec-2024). The model marries LSTM sequence learning with a multi-head self-attention layer and a post-attention gating block; a genetic algorithm tunes key hyperparameters (learning rate, hidden size, epochs). Using a 30-day horizon and standard preprocessing with lagged features, we benchmark SAG-LSTM against SA-LSTM and vanilla LSTM on MSE, RMSE, MAE, andR2, supplemented by error-trend and residual diagnostics, a forecast coherence score, and inter-asset dynamic/cross-correlation analyses. SAG-LSTM consistently dominates the baselines across assets: out-of-sampleR2rises to 0.90 for oil, 0.94 for gold, and 0.88 for Bitcoin, with visibly flatter error profiles and tighter, near-zero residuals. Inter-asset analyses show time-varying contemporaneous correlations but weak lead–lag effects, clarifying when co-movement is episodic rather than persistent. The largest gains occur in oil, reflecting more structured fundamentals; improvements for gold and Bitcoin are material but tempered by regime shifts and sentiment-driven jumps. Training time is higher due to GA search (≈2,121 s), but inference is fast (≈0.40 s), making the approach suitable for infrequent retraining with near-real-time scoring. Our findings highlight the value of hybrid, optimization-aware deep architectures for medium-horizon forecasting while underscoring the limits of price-volume inputs in sentiment-sensitive markets. These results offer actionable guidance for practitioners and a roadmap for future research and policy.
The rapid adoption of blockchain technology and generative AI contributes significantly to global electricity consumption, raising concerns about environmental sustainability. The first step in saving energy is to identify current consumption. However, since blockchain and generative AI are cloud-based services, it is difficult to understand electricity consumption outside one’s facilities. This creates a barrier for user companies and organizations seeking to increase the accuracy of calculating Scope 3 emissions. This study quantifies the electricity consumption of these technologies at a system-wide and per-use level. It compares them to traditional services such as payment networks and web search engines. Bitcoin, a Proof of Work (PoW) blockchain, consumes approximately 121 TWh, equivalent to 0.43% of global electricity consumption, and its energy demand per transaction is 720,000 times higher than that of the Visa payment system. Ethereum’s move to Proof of Stake (PoS) in 2022 reduces energy consumption by 99.988%, demonstrating the potential for efficiency gains. Generative AI models also have significant energy requirements, especially during the training and inference phases. For example, training GPT-4 required approximately 9450 MWh, and daily inference work exceeded 500 MWh. The results show that inference, driven by frequent user interaction, often exceeds the energy consumption of training. The study underscores the urgency of addressing these technologies’ environmental impact through strategies such as adopting energy-efficient consensus mechanisms and optimizing AI’s lifecycle. These findings are intended to guide organizations in refining their Scope 3 emissions calculations and adopting sustainable technology practices.
Abstract This study provides a comprehensive review of machine learning (ML) applications in the fields of business and finance. First, it introduces the most commonly used ML techniques and explores their diverse applications in marketing, stock analysis, demand forecasting, and energy marketing. In particular, this review critically analyzes over 100 articles and reveals a strong inclination toward deep learning techniques, such as deep neural, convolutional neural, and recurrent neural networks, which have garnered immense popularity in financial contexts owing to their remarkable performance. This review shows that ML techniques, particularly deep learning, demonstrate substantial potential for enhancing business decision-making processes and achieving more accurate and efficient predictions of financial outcomes. In particular, ML techniques exhibit promising research prospects in cryptocurrencies, financial crime detection, and marketing, underscoring the extensive opportunities in these areas. However, some limitations regarding ML applications in the business and finance domains remain, including issues related to linguistic information processes, interpretability, data quality, generalization, and the oversights related to social networks and causal relationships. Thus, addressing these challenges is a promising avenue for future research.