基于组合模型的新能源汽车销量预测方法

New Energy Vehicle Sales Forecasting Method Based on Combination Model

在汽车行业,企业需要对汽车的销量进行准确预测,以制定科学合理的生产计划。为此,以新能源汽车为例,利用历史销量数据构建基于传统时间序列理论的ARIMA预测模型和基于深度学习的LSTM预测模型,并利用Stacking方法对以上两种模型进行集成。实验结果表明,组合模型的平均绝对百分比误差为3.65%,相较于ARIMA(0,1,1)模型和LSTM模型分别降低2.57%和1.86%。该方法可为新能源汽车生产计划提供数据参考,并适用于其他类型汽车。

In view of the problem that enterprises need to accurately predict the sales volume of vehicles in order to formulate a scientific and reasonable production plan in the automotive industry, taking new energy vehicles as an example, using the historical sales volume data to build a ARIMA prediction model based on traditional time series theory and an LSTM prediction model based on deep learning, and uses the Stacking method to integrate the above two models. The experimental results show that the average absolute percentage error of the combined model is 3.65%, reduced by 2.57% and 1.86% respectively,compared with ARIMA(0,1,1) model and LSTM model. This method can provide data reference for the production plan of new energy vehicles, and is applicable to other type of car.