Cotton Futures Price Forecasting Method Based on LSTM
Traditional statistical and econometric time series forecasting models have limitations in complex financial markets.Long Short Term Memory(LSTM)networks in deep learning are believed to overcome these limitations.This study constructs a multi-layer LSTM network prediction model using cotton futures price data from 2019 to 2024 in China.The results indicate that adjusting the parameters of the LSTM network model significantly optimizes the prediction performance,especially in terms of the number of iterations,learning rate,window size,and network layers.Compared to K-nearest neighbor algorithm(KNN),multiple linear regression(MLR),and support vector regression(SVR)models,LSTM networks have higher prediction accuracy.Measured by the Mean Absolute Percentage Error(MAPE),the LSTM network reduces errors by 89.28%,85.92%,and 17.8%compared to KNN,MLR,and SVR models.The research results indicate that LSTM networks perform well in price prediction,providing new ideas for cotton futures price prediction.
万方数据
维普
