Jul 2026· International Journal of Advanced Research· Vol 14, pp. 896-904· 0 citations
TL;DR
A comparative analysis of machine learning and deep learning models for cryptocurrency price forecasting using historical Aave (AAVE) cryptocurrency data indicates that deep learning models, especially LSTM, are more suitable for capturing temporal dependencies and nonlinear patterns in cryptocurrency price data.
Abstract
Cryptocurrency markets are highly volatile, nonlinear, and affected by several internal and external market factors, making price forecasting a challenging task. Accurate cryptocurrency price forecasting can support investors, traders, and financial analysts in making informed decisions. This research paper presents a comparative analysis of machine learning and deep learning models for cryptocurrency price forecasting using historical Aave (AAVE) cryptocurrency data. The dataset consists of 275 records and 10 features, including Date, High, Low, Open, Close, Volume, and Marketcap. The Close price is selected as the target variable, while High, Low, Open, Volume, and Marketcap are used as predictor variables. Five models are implemented and compared: Linear Regression, Support Vector Regression, Random Forest Regressor, XGBoost Regressor, and Long Short-Term Memory. The models are evaluated using Mean Absolute Error, Root Mean Square Error, Mean Absolute Percentage Error, R-squared score, and directional accuracy. Experimental results show that the LSTM model achieved the best performance with the lowest RMSE of 2.74, MAE of 1.78, MAPE of 3.91%, and R-squared score of 0.965. The results indicate that deep learning models, especially LSTM, are more suitable for capturing temporal dependencies and nonlinear patterns in cryptocurrency price data.
The high volatility and complexity of cryptocurrency markets create difficulties for investors and researchers who
attempt to make accurate price predictions. This project presents an intelligent and data-driven approach for
cryptocurrency analysis using Long ShortTerm Memory (LSTM) networks, a specialized type of Recurrent Neural Network
(RNN) which enables the model to learn temporal dependencies across sequential data. The system builds predictive models
through training which uses historical price data that includes opening price, closing price, high, low, and trading volume.
The proposed model follows a structured workflow that includes data collection from financial APIs, preprocessing
techniques such as normalization and time-series windowing, and model training using LSTM architecture. The LSTM
model enables accurate predictions through its ability to learn long-term dependencies and patterns which exist in
cryptocurrency price movements. The evaluation process uses Root Mean Square Error (RMSE) and Mean Absolute Error
(MAE) metrics to test prediction accuracy. The system includes a core function which displays actual versus predicted price
data through graphical visualizations to assist users in understanding market patterns and model effectiveness. The
approach provides better interpretability which helps traders and analysts to make better decisions. The system supports
scalability which allows its application to various cryptocurrencies including Bitcoin and Ethereum and other digital
currencies.
D. Rahul, M. Shiva, Parag Ravikant Kaveri et al.· International Journal of Inn...· 0 citations
This study aimed to develop and evaluate a hybrid deep ensemble learning framework integrating CNN-LSTM, GRU, and Transformer architectures through stacking for accurate next-day cryptocurrency price prediction. This quantitative predictive study analyzed daily market data for five major cryptocurrencies, including Bitcoin, Ethereum, Binance Coin, Ripple, and Cardano, over the period from January 1, 2018, to December 31, 2025. The dataset comprised 14,610 cryptocurrency-day observations, with 2,922 observations for each asset. Data were chronologically divided into training, validation, and testing subsets. Historical price, trading volume, return, volatility, lagged variables, and technical indicators were used as predictors. A 30-day sliding window was employed to predict the following day’s closing price. The proposed framework combined CNN-LSTM, GRU, and Transformer models using a stacking-based meta-learner. Performance was evaluated using Mean Absolute Error, Root Mean Squared Error, Mean Absolute Percentage Error, and R². Diebold-Mariano tests were used to compare out-of-sample forecast accuracy. The proposed hybrid ensemble achieved an MAE of 0.0294, RMSE of 0.0436, MAPE of 3.12%, and R² of 0.971, outperforming CNN-LSTM, GRU, Transformer, ARIMA, Random Forest, XGBoost, and persistence benchmarks. Diebold-Mariano tests showed that the ensemble produced significantly lower forecast loss than all competing models, including Transformer, the strongest standalone model, with all p values below 0.001. Cryptocurrency-specific analyses yielded R² values above 0.960 for all five assets, while volatility-regime analysis showed that predictive accuracy declined under high-volatility conditions but remained strong, with an R² of 0.952. The findings demonstrate that combining convolutional, recurrent, and attention-based deep-learning mechanisms through stacking can significantly improve the accuracy and robustness of cryptocurrency price forecasting across heterogeneous assets and changing volatility conditions.
Mohammadreza Haghighi, Seyed Yashar Banihashem, Mohammadmehdi Gilaniansadeghi· Journal of Management and Bu...· 0 citations
A methodological framework based on LSTM neural networks for comparing the performance of various sets of features in predicting Bitcoin prices and proof that technical indicators have a significant impact on the predictive accuracy of this model, while social media sentiment has a minor impact under the circumstances are brought.
Sedeeq Hasan Banna, Ammar Ahmed Othman, Khaled Al-Raddah· SISTEMASI· 0 citations
This work attempts to provide a thorough comparative analysis mapping the precise accuracy–efficiency trade-off between Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models under a standardized Grid Search hyperparameter optimization pipeline using a recent Bitcoin closing-price dataset.
The paper examines the volatility dynamics of Bitcoin, Ether, XRP, and Solana as an indicator of market risk. Cryptocurrency price series are variable and non-stationary, which limits the accuracy of risk forecasting. The methodology combines long-memory estimation, multifractal analysis, and neural network modelling. Using daily data, returns, absolute returns, and 30-day rolling volatility were calculated. Bitcoin showed the lowest average volatility, 0.0326, whereas Solana showed the highest, 0.0554; Ether and XRP reached 0.0428 and 0.0458, respectively. The long-memory parameter of price series ranges from 0.4788 to 0.5420, while volatility series exhibit moderate long memory of 0.2326–0.2964 and confirmed stationarity. XRP showed the widest multifractal volatility spectrum, 1.2651. A distinctive feature of the proposed approach is the use of LSTM models with market, lagged, and dynamic fractal features constructed within a rolling window. Forecasting was performed for 1-, 3-, 7-, and 14-day horizons, and the results were aggregated over 12 repeated training runs. Comparison of LSTM models with and without fractal features showed that their forecasting contribution depends on the feature construction method, while volatility-based features were generally more informative. Comparison with GARCH, ARFIMA, and HAR showed that performance depends on the asset, horizon, and evaluation metric.
Yaroslav Sokolovskyy, M. Opryshko, Tetiana Samotii et al.· Applied Sciences· 0 citations
In financial time series markets, Bitcoin price forecasting is challenging due to the high nonlinearity, volatility, nonstationarity, and noise characterizing cryptocurrency markets. In this paper, we propose a novel hybrid linear/nonlinear forecasting model based on a kernel regression decomposition framework, which exploits nonparametric kernel smoothing methods and linear/nonlinear forecasting models to improve forecast accuracy. The Bitcoin price series is decomposed into the long-term trend and short-term fluctuation components using four kernel functions: Uniform, Gaussian, truncated Gaussian, and Epanechnikov. The decomposed components are modeled separately by linear models (AR and ARIMA) and nonlinear models (NPAR and NNAR), resulting in sixty-four hybrid forecasting combinations. The framework is tested on BTC/USD, BTC/GBP, BTC/EUR, BTC/CNY, and BTC/JPY markets through one-step-ahead forecasting, using different performance metrics and a statistical forecasting test. The empirical results reveal that the proposed decomposition-based hybrid models outperform the best forecasting models proposed in the literature, both single and direct-hybrid models, in all markets. The results confirm improved accuracy, robustness, and predictive reliability in forecasting cryptocurrency markets based on the proposed kernel-based decomposition.
Hasnain Iftikhar, Said Farooq Shah, Paulo Canas Rodrigues et al.· IEEE Access· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
