Paper: arXiv 2510.15900
Authors: Emmanuel Boadi
Abstract
This study proposes a hybrid deep learning model for forecasting the price of Bitcoin, as the digital currency is known to exhibit frequent fluctuations. The models used are the Variational Mode Decomposition (VMD) and the Long Short-Term Memory (LSTM) network. First, VMD is used to decompose the original Bitcoin price series into Intrinsic Mode Functions (IMFs). Each IMF is then modeled using an LSTM network to capture temporal patterns more effectively. The individual forecasts from the IMFs are aggregated to produce the final prediction of the original Bitcoin Price Series. To determine the prediction power of the proposed hybrid model, a comparative analysis was conducted against the standard LSTM. The results confirmed that the hybrid VMD+LSTM model outperforms the standard LSTM across all the evaluation metrics, including RMSE, MAE and R2 and also provides a reliable 30-day forecast.
Complexity vs Empirical Score
- Math Complexity: 8.5/10
- Empirical Rigor: 6.5/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced mathematical concepts such as variational mode decomposition (VMD) with detailed variational formulations and LSTM gate equations, scoring high in math complexity. However, while it presents specific backtesting results (RMSE, MAE, R2) on historical data, it lacks code, dataset access, or implementation details, and the methodology is standard for deep learning research, placing it in the Lab Rats quadrant.
Research Flowchart
flowchart TD
A["Research Goal<br>Bitcoin Price Forecasting"] --> B{"Data Input:<br>Historical Bitcoin Price Data"}
B --> C["Methodology: Decomposition<br>Variational Mode Decomposition VMD"]
C --> D["Process: Modeling<br>Long Short-Term Memory LSTM Network"]
D --> E["Outcome: Forecasting<br>30-Day Bitcoin Price Prediction"]
E --> F["Result: Comparison<br>VMD+LSTM vs Standard LSTM"]
F --> G["Conclusion<br>Hybrid Model Outperforms Standard LSTM<br>RMSE, MAE, R2 Metrics"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a hybrid deep learning model that combines Variational Mode Decomposition (VMD) with a Long Short-Term Memory (LSTM) network to forecast Bitcoin prices.
Source sentence
This study proposes a hybrid deep learning model for forecasting the price of Bitcoin, as the digital currency is known to exhibit frequent fluctuations.
Assumes 5
- The analysis uses daily closing prices of Bitcoin as the target series. data
Source sentence
The dataset consists of daily Bitcoin closing prices obtained from Yahoo Finance [? ] from August 11, 2017, to June 13, 2025.
- Missing values are removed and the series is normalized with MinMax scaling before modeling. data
Source sentence
After downloading, missing values were removed using the dropna() method... The
- The VMD decomposition parameter K is chosen based on residual energy, with K=10 selected as optimal. model
Source sentence
To determine the optimal number of modes, the residual energy ratio was computed for values of K ranging from 5 to 20
- Forecast performance is evaluated using standard regression metrics (RMSE, MAE, MSE, R2). statistical
Source sentence
The performance of each model was evaluated using standard regression metrics.
Show 1 more
- The study assumes that forecasting accuracy on historical data reflects future predictive ability (no explicit out‑of‑sample trading simulation). other
Source sentence
The results demonstrate that both models are capable of forecasting Bitcoin prices, with the VMD-LSTM model achieving superior performance.
Data
- Daily Bitcoin closing prices market-prices public
Source sentence
The dataset consists of daily Bitcoin closing prices obtained from Yahoo Finance [? ] from August 11, 2017, to June 13, 2025. A total of 2,863 daily observations were recorded.
Costs
Not applicable
Finds 2
- The hybrid VMD+LSTM model outperforms the plain LSTM across all evaluation metrics, achieving RMSE 1619.63 vs 4005.32, MAE 1397.18 vs 3157.91, MSE 2.62 million vs 16.04 million, and R² 0.9934 vs 0.9598 on the test set. performance
Source sentence
The results show that the VMD+LSTM model outperformed the plain LSTM on both training and test sets. Notably, the RMSE and MAE values for the VMD+LSTM model on the test set were significantly lower (RMSE: 1619.63 vs. 4005.32; MAE: 1397.18 vs. 3157.91)
- The hybrid model consistently yields lower prediction errors and higher R², indicating more accurate and reliable 30‑day Bitcoin price forecasts. predictive-accuracy
Source sentence
The hybrid VMD+LSTM model outperforms the standard LSTM across all the evaluation metrics, including RMSE, MAE and R2 and also provides a reliable 30-day forecast.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 11 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (15)
- A Bayesian regularized neural network for analyzing bitcoin trends (2021)
- Analysis of Blockchain technology: pros, cons and SWOT (2019)
- Bitcoin in the economics and finance literature: a survey (2021)
- Comparitive automated bitcoin trading strategies (2016)
- Comparison of arima time series model and lstm deep learning algorithm for bitcoin price forecasting (2018)
- Variational mode decomposition (2013)
- Bayesian regression and Bitcoin (2014)
- Bitcoin price prediction and analysis using deep learning models (2020)
and 7 more
- Machine learning algorithm for cryptocurrencies price prediction (2021)
- Bitcoin price forecasting with neuro-fuzzy techniques (2019)
- Predicting the price of bitcoin using machine learning (2018)
- An Intelligent Predictive Maintenance Framework with Autonomous Feature Selection based on Hybrid Fuzzy Set and Rough Set Theories for Multiclass Fault Classification (2022)
- A comparative study of bitcoin price prediction using deep learning (2019)
- A GARCH-MIDAS approach to modelling stock returns (2024)
- AI meets economics: Can deep learning surpass machine learning and traditional statistical models in inflation time series forecasting? (2025)
0 of 15 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.