Is it difficult to predict the price movements of high-volatility assets

Predicting the price movements of assets with high volatility is typically considered challenging. However, the hybrid model we propose demonstrates that, for such assets, forecasting the direction of price movement is not significantly more difficult than in low-volatility contexts. Rather, the key challenge lies only in the level accuracy of predicted price. Nevertheless, for trading purposes, the direction of price movement is often more crucial. To capture mixed patterns in price sequences, we first decompose the original time series into several mode components by Time Varying Filtering based Empirical Mode Decomposition (TVF-EMD). Then we employ Deep Extreme Learning Machines (DELM) to model the nonlinearity of the prediction function and utilizes the state-of-the-art Marine Predators Algorithm (MPA), a swarm intelligence algorithm, to optimize both the intrinsic parameters and hyperparameters of the deep network. Moreover, by incorporating a specially designed fitness function that accounts for directional errors, we significantly enhance price trend prediction in high-volatility scenarios. Finally, the trading strategy derived from our hybrid model yields a 69.10% more profit in the back-testing of weekly prediction than the Buy & Hold strategy.