A novel cryptocurrency price time series hybrid prediction model via machine learning with MATLAB/Simulink

摘要

Bitcoin is widely recognized as the first decentralized digital cryptocurrency based on blockchain technology. Its unique properties make it a leading contender in the realm of digital currencies, and it continues to maintain a dominant position in the short-term. However, the high volatility of bitcoin prices poses significant challenges for prediction models. Therefore, it is important to develop rational processing techniques to weaken the volatility of raw data, thereby facilitating more accurate predictions. To this end, we propose a novel time series hybrid prediction model (TSHPM) to estimate bitcoin prices. Our approach utilizes variational mode decomposition (VMD) to decompose daily bitcoin prices into several simple modes. We then use approximate entropy (ApEn) for modal characterization and sequence reconstruction to determine the complexity of the different components of the time series. To better compare the accuracy of different models, we establish a comprehensive evaluation index (CEI). Furthermore, we adopt an innovative approach by utilizing the Simulink module in MATLAB for machine learning prediction. Through a rigorous selection process, we identify the appropriate model that produces the best prediction. Among them, the VMD-LSTM-Adam model has the best prediction performance with a CEI value of only 0.4017. Empirical analysis demonstrates that our TSHPM approach significantly outperforms traditional prediction methods, reducing prediction errors by more than 50. At the same time, the time complexity of the prediction is optimized by about 15 and the overall performance of the model is greatly improved. In summary, our findings demonstrate the effectiveness of the TSHPM model in predicting complex time series, particularly bitcoin prices. Our approach provides a promising avenue for further research in the field of cryptocurrency price prediction, with the potential to facilitate more accurate and reliable predictions in future.