基于长短时记忆神经网络的 生猪价格预测模型.

Following pseudo-periodic time series, the 11 cycles 11 of hog price repeats differently in length and shape, which causes difficulties in prediction. To solve the problem, a hog price prediction method was proposed based on long short term memory ( LSTM ) network model. The prediction factors were preprocessed and selected by PCA model, and the key parameters of LSTM model were optimized by firefly algorithm (FA) . According to the parameters, the networks were trained to establish 3 prediction models, and the models were used to predict hog prices of the following 1, 2 and 8 weeks. The results show that according to the proposed method, the mean absolute error, the root mean squared error and the decision coefficient are respective 1. 455 8, 4. 910 2 and 92. 57%, which illuminates that compared with conventional shallow prediction models and unoptimized LSTM model, the proposed method exhibits superior performance. The proposed method can deal with the changes of the cycles of hog price and is suitable for the prediction of hog price and other price series with the same characteristics. [ABSTRACT FROM AUTHOR]