Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Bitcoin Price Forecasting Based on Hybrid Variational Mode Decomposition and Long Short Term Memory Network

Emmanuel Boadi

TL;DR

This paper tackles the volatility and nonstationarity of Bitcoin prices by introducing a hybrid Variational Mode Decomposition (VMD) and Long Short-Term Memory (LSTM) framework. By decomposing the price series into 15 Intrinsic Mode Functions (IMFs) with VMD and training a separate LSTM on each IMF, the method ensembles the IMF forecasts to produce a final price prediction, improving accuracy over a plain LSTM on raw data. Empirical results using 2,863 daily observations (2017-2025) show superior performance across RMSE, MAE, MSE, and $R^2$, and the approach yields reliable 30-day forecasts, suggesting strong potential for cryptocurrency forecasting and other nonstationary domains. The combination of signal denoising, frequency separation, and per-component modeling provides meaningful improvements in capturing both short- and long-term dynamics while mitigating overfitting.

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.

Bitcoin Price Forecasting Based on Hybrid Variational Mode Decomposition and Long Short Term Memory Network

TL;DR

This paper tackles the volatility and nonstationarity of Bitcoin prices by introducing a hybrid Variational Mode Decomposition (VMD) and Long Short-Term Memory (LSTM) framework. By decomposing the price series into 15 Intrinsic Mode Functions (IMFs) with VMD and training a separate LSTM on each IMF, the method ensembles the IMF forecasts to produce a final price prediction, improving accuracy over a plain LSTM on raw data. Empirical results using 2,863 daily observations (2017-2025) show superior performance across RMSE, MAE, MSE, and , and the approach yields reliable 30-day forecasts, suggesting strong potential for cryptocurrency forecasting and other nonstationary domains. The combination of signal denoising, frequency separation, and per-component modeling provides meaningful improvements in capturing both short- and long-term dynamics while mitigating overfitting.

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.
Paper Structure (19 sections, 4 equations, 6 figures, 2 tables)

This paper contains 19 sections, 4 equations, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Overview
  5. Data
  6. Variational Mode Decomposition (VMD)
  7. Long Short-Term Memory (LSTM) Network
  8. Long Short-Term Memory (LSTM) Network
  9. Hybrid VMD-LSTM Forecasting Framework
  10. Model Evaluation
  11. Results and Discussion
  12. Overview
  13. Descriptive Analysis
  14. Residual Energy and Mode Selection (K)
  15. IMF Decomposition Results
  16. ...and 4 more sections

Figures (6)

  • Figure 1: Residual Energy vs. Number of Modes (K)
  • Figure 2: Selected IMF Components from VMD Decomposition of Bitcoin Price
  • Figure 3: Learning Curves for IMF 2, 7, and 15
  • Figure 4: Actual vs. Predicted Bitcoin Prices (VMD+LSTM)
  • Figure 5: 30-Day Bitcoin Price Forecast using VMD-LSTM
  • ...and 1 more figures