The emergence of Bitcoin as a pioneering cryptocurrency has transformed financial markets, garnering widespread interest from academicians, policymakers, and investors. The market's inherent volatility and the rapid integration of public information into price movements continue to present a formidable challenge in accurately forecasting Bitcoin prices despite its potential. The limitations of conventional financial models, which frequently need to consider the distinctive attributes of cryptocurrencies, further exacerbate this challenge. Despite the proliferation of ML in various fields, existing models have not fully harnessed these techniques, performing only marginally better than random guesses due to the unique challenges posed by the high volatility and complex dynamics of cryptocurrency markets. This study introduces a systematic review of ML methods specifically tailored for Bitcoin price prediction, with a focus on evaluating the robustness, accuracy, and appropriateness of advanced ML techniques like Long Short-Term Memory (LSTM) networks. The novelty lies in its comprehensive assessment of these methods in the context of data-driven marketing, aiming to enhance both academic understanding and practical applications in financial technology. The previous studies haven't Machine Learning (ML) has become a formidable instrument that has the potential to improve the accuracy of forecasting; however, there still needs to be more comprehension regarding the most effective ML models in this field. The study's importance is derived from its systematic examination of various machine learning (ML) techniques employed to predict the price of Bitcoin, with a particular emphasis on their integration into data-driven marketing strategies. The results will substantially contribute to both academic research and practical applications, providing valuable insights that can be used to develop more dependable forecasting tools, thereby benefiting investors, marketers, and policymakers.
Yang Zhou, Chi Xie, Gang‐Jin Wang, Jue Gong · 5 authors
Abstract Cryptocurrency is a remarkable financial innovation that has affected the financial system in fundamental ways. Its increasingly complex interactions with the conventional financial market make precisely forecasting its volatility increasingly challenging. To this end, we propose a novel framework based on the evolving multiscale graph neural network (EMGNN). Specifically, we embed a graph that depicts the interactions between the cryptocurrency and conventional financial markets into the predictive process. Furthermore, we employ hierarchical evolving graph structure learners to model the dynamic and scale-specific interactions. We also evaluate our framework’s robustness and discuss its interpretability by extracting the learned graph structure. The empirical results show that (i) cryptocurrency volatility is not isolated from the conventional market, and the embedded graph can provide effective information for prediction; (ii) the EMGNN-based forecasting framework generally yields outstanding and robust performance in terms of multiple volatility estimators, cryptocurrency samples, forecasting horizons, and evaluation criteria; and (iii) the graph structure in the predictive process varies over time and scales and is well captured by our framework. Overall, our work provides new insights into risk management for market participants and into policy formulation for authorities.
N. Hafidi, Zakaria Khoudi, Mourad Nachaoui, Soufiane Lyaqini
Accurate cryptocurrency price forecasting is crucial for investors and researchers in the dynamic and unpredictable cryptocurrency market. Existing models face challenges in incorporating various cryptocurrencies and determining the most effective hyperparameters, leading to reduced forecast accuracy. This study introduces an innovative approach that automates hyperparameter selection, improving accuracy by uncovering complex interconnections among cryptocurrencies. Our methodology leverages deep learning techniques, particularly Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) networks, in conjunction with the Genetic Algorithm (GA) to optimize hyperparameters. We propose and compare two architectures, LAO and LOEE, utilizing these methods to enhance forecast accuracy and address the challenges of the cryptocurrency market. This cutting-edge approach not only improves forecasting capabilities but also provides valuable insights for managing cryptocurrency investments and conducting research. By automating hyperparameter selection and considering interconnections between cryptocurrencies, our approach offers a practical solution for accurate cryptocurrency price prediction in a dynamic market environment, benefiting both investors and academics.
V. G. Murugan, Sudhansu Sekhar Nanda, V. R. Sudindra, Gajendra Naidu Jatty · 6 authors
Financial portfolio management receives secure and efficient processing through the combination of Federated Learning (FL) with blockchain technology. FL protects data privacy through its distributed training capabilities that prevent financial data exposure. Blockchain secures operations and maintains transparency through the creation of unalterable model change logs on its ledger systems. A proposed approach reaches 91.3% model accuracy level and a 1.27 Sharpe Ratio and delivers a 14.3% return on investment which surpasses conventional approaches. The training duration reaches 4.8 hours while security enhancements and increased efficiency become available in the system. The designed framework supports 200 transactions per second at 250 ms response time while maintaining complete security with 0.10 USD transaction fees. It demonstrates the ability of FL and blockchain technology to boost security measures and risk-adjusted performance metrics and computational capability. Future Work will concentrate on blockchain performance optimization while testing practical applications of algorithms and revising algorithms to optimize financial applications.
Our research investigates the predictive performance and robustness of machine learning classification models and technical indicators for algorithmic trading in the volatile cryptocurrency market. The main aim is to identify reliable approaches for informed decision-making and profitable strategy development. With the increasing global adoption of cryptocurrency, robust trading models are essential for navigating its unique challenges and seizing investment opportunities. This study contributes to the field by offering a novel comparison of models, including logistic regression, random forest, and gradient boosting, under different data configurations and resampling techniques to address class imbalance. Historical data from cryptocurrency exchanges and data aggregators is collected, preprocessed, and used to train and evaluate these models. The impact of class imbalance, resampling techniques, and hyperparameter tuning on model performance is investigated. By analyzing historical cryptocurrency data, the methodology emphasizes hyperparameter tuning and backtesting, ensuring realistic model assessment. Results highlight the importance of addressing class imbalance and identify consistently outperforming models such as random forest, XGBoost, and gradient boosting. Our findings demonstrate that these models outperform others, indicating promising avenues for future research, particularly in sentiment analysis, reinforcement learning, and deep learning. This study provides valuable guidance for navigating the complex landscape of algorithmic trading in cryptocurrencies. By leveraging the findings and recommendations presented, practitioners can develop more robust and profitable trading strategies tailored to the unique characteristics of this emerging market.
Non-fungible tokens (NFTs) have gained mainstream attention in the fintech community, but there is little research on their statistical properties. This study investigates the long-memory characteristics of NFT returns and volatility, focusing on their potential for predicting price movements. As NFTs do not conform to traditional models, understanding their unique features is crucial for comprehending complex market dynamics. This study aims to reveal the impact of macroeconomic factors on NFT prices, understand their correlation and develop predictive models using autoregression and artificial intelligence (AI) technology. This research utilized datasets from the Centers for Disease Control and Prevention (CDC), U.S. Bureau of Labor Statistics, Bureau of Economic Analysis, Christie’s, Dune, and Google Trends. Correlation and p value tests revealed strong relationships between NFT prices and variables such as weekly volume, pandemics, inflation and security. The Baseline Model using autoregression with NFT volume, security and technology factors outperformed all other models demonstrating the speculative volatility of NFTs. The Transformer Model using transformers, an architecture used by ChatGPT, Gemini and Stable Diffusion, showed high accuracy with less feature selection and preprocessing efforts. This study provides a novelty using a systematic approach for researchers to perform financial forecasting and contributes to the scarce literature on NFTs. This research offers valuable insights to investors and private agents regarding the right economic conditions for NFT investments by reducing portfolio risks and making informed decisions. To the authors’ best knowledge, this is the first study to utilize time-series transformers for forecasting NFTs based on macroeconomic factors.
This study explores the integration of technical indicators, specifically Exponential Moving Averages (EMA) and Volume Weighted Average Price (VWAP), into machine learning models for cryptocurrency price forecasting. Our findings reveal that including these indicators can complicate the modeling process without necessarily improving performance. Support Vector Regression (SVR) and Random Forest Regressor (RFR) models outperform deep learning approaches such as Long Short-Term Memory (LSTM) networks, demonstrating higher predictive accuracy with simpler feature sets. These findings emphasize the challenges of high-dimensional data and the critical role of rigorous feature selection and preprocessing in financial forecasting. Practical implications and trade-offs between model complexity and prediction accuracy are discussed, providing valuable insights for researchers and practitioners in financial analytics.
<p>Blockchain technology is the foundation of cryptocurrencies, which are virtual currencies. The decentralized nature of cryptocurrencies has resulted in a significant reduction of central authority over them, which has implications for global trade and relations. The need for an effective model to anticipate the price of cryptocurrencies is essential due to their wide variations in value. Due to the shortcomings of conventional production forecasting, in this work, four distinct models were used. The deep learning models are the long short-term memory (LSTM) and bidirectional long short-term memory (Bi-LSTM), and both the Facebook-Prophet and Silverkite support the exponential smoothing technique. Silverkite is designed to handle a wide range of time series forecasting tasks. Considering past bitcoin information from January 2012 to March 2021, a period of nine years, we looked at the models. The Bi-LSTM model yields a 7.073 mean absolute error (MAE) and a 3.639 root mean squared error (RMSE). The Bi-LSTM model identifies the deviations that might draw attention and avert any problems.</p>
Muhamad Abdul Fatah, Martanto Martanto, Arif Rinaldi Dikananda, Ahmad Rifa’i
Predicting erratic assets like Ethereum is difficult in the dynamic cryptocurrency market. This study uses an enhanced Support Vector Regression (SVR) algorithm to create a daily price prediction model for Ethereum. Yahoo Finance provided the data, which was preprocessed to include missing value cleaning, normalization, and feature extraction of Moving Average (MA) and Exponential Moving Average (EMA). The data was collected between August 4, 2019 and August 4, 2024. An ideal combination was obtained by parameter optimization with GridSearchCV: gamma scale, linear kernel, epsilon of 1, and C of 100. The model performed well, as evidenced by its R2 of 0.9985 and MSE of 2137.97. The model's reliability in predicting Ethereum's price movement patterns was validated via prediction graphs. A 30-day forecast indicated a stable trend, with prices slightly decreasing from $2921.31 on January 1, 2025, to $2919.83 on January 31, 2025. These results highlight the importance of data preprocessing and parameter optimization in enhancing SVR model performance.
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.
Abhishek Choubeya, Dwi Harini, Y. Ishwarya, J. Shravya · 5 authors
As the first decentralized cryptocurrency, Bitcoin is a type of digital money. Bitcoin transactions between users do not require a centralized authority, such as banks or governments. It conducts transactions using blockchain technology. Bitcoin is an open-source project that is not restricted to any one authority, and anybody can participate in creating bitcoins. Bitcoin was outlawed in India for allegedly facilitating money laundering and being used illegally. The paper we present here will explore how machine learning algorithms could be used to forecast Bitcoin values. A variety of time-series analysis models, such as AR, ARMA, ARIMA, and SARIMAX, and regression analysis models, such as Bayesian, polynomial, and Elastic Net, are used in machine learning algorithms to forecast the prices of Bitcoin.
Hani Alnami, Muhammad Mohzary, Basem Assiri, Hussein Zangoti
The accurate prediction of cryptocurrency prices is crucial due to the volatility and complexity of digital asset markets, which pose significant challenges to traders, investors, and researchers. This research addresses these challenges by leveraging machine learning and deep learning techniques to forecast closing prices for cryptocurrencies, focusing on Bitcoin, Ethereum, Binance Coin, and Litecoin cryptocurrency datasets. A Random Forest ensemble learning algorithm, a Gradient Boosting model, and a feedforward neural network were implemented to handle the complexities in cryptocurrency data. A Z-Score-based anomaly detection framework was integrated to classify closing prices as normal or abnormal, aiding in identifying significant market events. Evaluation metrics, such as the Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-squared (R2), demonstrate the superior precision and reliability of the Random Forest and Gradient Boosting models. The deep learning model indicates strong generalization capabilities, suggesting potential advantages on more complex datasets. These findings highlight the importance of combining advanced machine learning techniques and cryptocurrencies to develop a robust framework for cryptocurrency forecasting and anomaly detection.
Bo Gao, Yuan Wang, Qingsong Wei, Yong Liu · 6 authors
Decentralized finance (DeFi) applications depend on accurate price oracles to ensure secure and fair transactions. However, poorly integrated oracles remain susceptible to manipulation, enabling attackers to exploit smart contract logic for unfair asset valuation and financial gain. While many such vulnerabilities are only detected after deployment, smart contracts are typically immutable once deployed, making post-hoc fixes costly or infeasible. This highlights the critical need for detecting oracle manipulation risks before deployment. In this paper, we propose$AiRacleX$, a novel LLM-driven framework that enables pre-deployment detection of price oracle manipulation vulnerabilities by leveraging the complementary strengths of multiple large language models (LLMs). Our approach begins with domain-specific knowledge extraction, where an LLM model synthesizes precise insights about price oracle vulnerabilities, eliminating the need for profound expertise from developers or auditors. This knowledge forms the foundation for a second LLM model to generate structured, context-aware Chain-of-Thought prompts, which guide a third LLM model in accurately identifying manipulation patterns in smart contracts. We evaluate$AiRacleX$on 60 known vulnerabilities from 44 real-world DeFi exploits and Code4rena projects spanning 2021-2023. The results show that$AiRacleX$achieves a 2.58 times improvement in recall over the state-of-the-art GPTScan, with comparable precision. Our framework also demonstrates strong extensibility and efficiency, and supports deployment with open-source LLMs to enhance security and reduce operational cost.
Forecasting bitcoin prices using deep learning techniques has gained significant attention recently. Despite the success of Generative Adversarial Networks (GANs) across various domains, there is a crucial issue in bitcoin price prediction: the ability to accurately capture correlations between temporal data points. Moreover, GANs face notable inherent challenges, particularly training instability and mode collapse. In this paper, we employ windowing and conditioning techniques to capture correlations between prices and other relevant features such as technical indicators. Additionally, we use a hybrid Long Short-Term Memory (LSTM)-based generator that combines a convolution layer and LSTM to improve learning from temporal data. Appropriate loss functions are also utilized for the discriminator and generator to enhance training stability and mitigate mode collapse. The proposed method was evaluated on the historical Bitcoin data sourced from the Yahoo Finance website. Experimental results demonstrate that the proposed method significantly outperforms state-of-the-art methods. The source code for the proposed method is available on GitHub: https://github.com/mahdimanian/draganbtc/
Bitcoin price prediction remains a complicated task in financial markets since price fluctuations are characterized by high variability, non-linear and dependent on different trends. The common models of forecasting can be disruptive because they do not readily accommodate the breadth of the data and the dynamics of the marketplace. The paper presents a new approach to predicting Bitcoin price shifts based on state-of-the-art data structures and both machine learning and technical analysis. The formative historical price datasets and large scale features are organized and manipulated in an efficient manner with the aid of data structures as well as the temporal and non-linear trends in prices are captured by ARIMA, Exponential smoothing and Random. Also, trends of moving averages are used to boost the model's competency to analyze the sentiments of the market. The proposed framework was put through a test on real world Bitcoin price data and it was shown to provide better results as compared to the traditional methods with better accuracy and visualization. Besides that, our method enhances the accuracy of the forecast, while the deployment of the method allows developing a realtime Bitcoin price prediction model that can be useful for traders and investors. The integration of machine learning and technical analysis gives a powerful and reliable tool to forecast the Bitcoins price trends.
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.
Bitcoin, the pioneering cryptocurrency, is renowned for its extreme volatility and speculative nature, making accurate price prediction a persistent challenge for investors. While recent studies have employed multivariate models to integrate historical price data with social media sentiment analysis, this study focuses on improving an existing univariate approach By incorporating sentiment and tweet volume data into a multivariate framework, we systematically evaluated the benefits of this integration. Among the five LSTM-based models developed for this study, the Multi-LSTM-Sentiment model achieved the best performance, with the lowest mean absolute error (MAE) of 0.00196 and root-mean-square error (RMSE) of 0.00304. These results underscore the significance of including social media sentiment in predictive modelling and demonstrate its potential to enhance decision-making in the highly dynamic cryptocurrency market.
Jahanzaib Alvi, Kehkashan Nizam, S. M. A. Jafri, Muhammad Rehan · 5 authors
This paper investigates the efficacy of deep learning models such as Long-Short Term Memory (LSTM), Convolutional Neural Networks (CNN), and Gated Recurrent Units (GRU) for cryptocurrency price prediction, examining their short-term and long-term forecasting accuracy for investor guidance and advancing AI in financial analysis. The study uses time series analysis with LSTM, CNN, and GRU models on daily cryptocurrency prices from Investing.com, preprocessing data before testing on Bitcoin, Ethereum Classic, Ethereum, Litecoin, Monero, and the other 37 cryptocurrencies. RMSE, MAE, and accuracy rates measure performance. Findings revealed that only six cryptocurrencies were selected for final analysis, including Bitcoin, Ethereum Classic, Ethereum, Litecoin, and Monero. Results indicate that the deep learning models, particularly the LSTM and GRU, can predict cryptocurrency prices with high accuracy, especially for short-term forecasts within a 7-day window. The CNN model demonstrates significant predictive power, suggesting its utility for immediate trading decisions. Across the models, short-term precision was remarkably high, while long-term predictions maintained a moderate level of accuracy. This study presents a comparative analysis of LSTM, GRU, and CNN models for forecasting cryptocurrency prices, emphasizing LSTM and GRU's ability to navigate price volatility and suggesting their use for real-time trading analysis. The study's historical data reliance curtails forecasting unforeseen market shifts. Future studies should include new variables like social sentiment and blockchain analytics and test real-time adaptive models to enhance predictive strength. Model validation in actual market conditions is recommended for practical application.
This research compares the effectiveness of the hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) method and the Convolutional Neural Network (CNN) method in predicting the Ethereum (ETH) exchange rate against the United States Dollar (USD). The research uses historical ETH/USD data from Yahoo Finance for the period 2017-2022. Evaluation of the two models was carried out using the performance metrics Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), coefficient of determination (R²), and accuracy rate. The results showed that the CNN-LSTM hybrid model significantly outperformed the CNN model in predicting the ETH/USD exchange rate with a Test RMSE value of 94.67 compared to 129.02 for CNN, as well as an accuracy rate of 96.31% versus 94.89%. These findings contribute to the fintech literature by providing empirical evidence of the superiority of hybrid methods for high volatility cryptocurrency exchange rate prediction.
Accurately forecasting price swings is nowadays essential to investors looking to maximize their portfolios as the cryptocurrency markets continue to develop and fluctuate rapidly. The intricate, non-linear patterns in these markets are sometimes difficult for traditional financial models to depict. In response, this paper presents two machine learning techniques for predicting bitcoin prices: Extreme Gradient Boosting and Long Short-Term Memory. The study first evaluates how well these models forecast Bitcoin prices, assessing their accuracy with measures like Mean Absolute Error and Root Mean Squared Error. Four significant cryptocurrencies are then predicted by LSTM. In order to allocate assets in a way that optimizes returns while reducing risk, the forecasted prices are then incorporated into portfolio optimization algorithms utilizing Monte Carlo simulation and the efficient frontier. The results of the study show how machine learning approaches may be used to improve investing strategies through optimal portfolio allocation, in addition to projecting cryptocurrency values.