基于组合预测模型的铁路货运量预测研究

为了进一步提高铁路货运量的预测精度,提出基于乘积季节模型与引入注意力机制(Attention Mechanism)的长短期记忆(Long Short-Term Memory)模型的组合预测模型。首先建立乘积季节模型、LSTM模型与引入注意力机制的LSTM模型,然后利用误差修正法分别将2种LSTM模型与乘积季节模型组合起来进行预测,最后将预测结果分别与单一模型进行对比。采用2005年至2018年全国铁路月度货运量进行预测分析,结果表明2种组合预测模型的预测精度均高于单一预测模型的预测精度,其中基于乘积季节模型与引入注意力机制的LSTM模型的组合预测模型精度最高,具有研究和实用价值。

Railway freight volume forecasting based on a combined model

In order to further improve the prediction accuracy of railway freight volume, this paper proposed a combined prediction model based on multiplicative seasonal ARIMA model and the LSTM(Long Short-Term Memory) model that introduced the attention mechanism. Firstly, a product seasonal model, an LSTM model and an LSTM model with attention mechanism were established. Then, two types of LSTM models were combined with the product seasonal model for prediction using the error correction method. Finally, the prediction results were compared with single model. Based on the analysis of the monthly railway freight volume from 2005 to 2018, the results show that the prediction accuracy of the two combined prediction models is higher than that of the single prediction model. Among them, the combined prediction model based on multiplicative seasonal ARIMA model and the LSTM model with attention mechanism has the highest accuracy, and the experiment has research and practical value.