Mode Decomposition and Hybrid Modeling in Bitcoin Price Prediction
The unique production,issuance,and trading mechanisms,among many other fac-tors,have led to extreme volatility in the Bitcoin price,leading to the complexity of the forecasting task.A hybrid prediction model based on Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(ICEEMDAN)is proposed,which decomposes a complex original sequence into multiple simple intrinsic mode functions(IMFs)and sets the IMFs into components of different fre-quencies by a reconstruction algorithm.According to the different data patterns of each component,dif-ferent machine learning models are selected to make predictions separately,and the prediction results of each component are superimposed to get the final bitcoin price prediction results.The comparison re-sults show that the model is better than the single prediction model in each evaluation index,and the hybrid model can optimize the prediction results and reduce the prediction error better.
万方数据
维普
