Futures Price Forecasting Based on the Feature Fusion LSTM Model Using Long-Term Price Patterns

Forecasting of futures price is of great significance to the research of financial products and the formulation of trading strategies. Compared with traditional econometric models, deep learning algorithm has outstanding advantages in the application of the financial field by virtue of its excellent nonlinear information processing ability. But few existing studies take long-term price patterns as input characteristics, so this paper proposes a novel deep learning model fusing long-term price model features to predict the closing price of futures. The clustering algorithm is served to extract features of long-term price patterns, and the deep learning component selects associated LSTM. Finally, the experimental results on several products in the Chinese futures market show that our method achieves better forecasting performance than other similar methods.