准确预测电子商务销售额对电子商务业高质量发展具有重大意义。本文以贵州省为例,基于收集到的5个预测指标,建立可加模型对贵州省电子商务销售额未来三年进行预测。首先使用变分自编码器(VAE)方法对2003~2012年电子商务销售额缺失值进...准确预测电子商务销售额对电子商务业高质量发展具有重大意义。本文以贵州省为例,基于收集到的5个预测指标,建立可加模型对贵州省电子商务销售额未来三年进行预测。首先使用变分自编码器(VAE)方法对2003~2012年电子商务销售额缺失值进行估计补充,然后使用互信息评估5个预测指标被选择的合理性,同时使用基于B样条的核密度回归估计方法对可加模型进行估计。最后使用GM (1, 1)模型预测未来三年5个预测指标值,然后带入到已建立的可加模型中,从而得到电子商务销售额未来三年的预测值。Accurate prediction of e-commerce sales is of significant importance for the high-quality development of the e-commerce industry. Taking Guizhou Province as an example, this paper establishes an additive model to forecast the e-commerce sales in Guizhou Province for the next three years based on five collected predictive indicators. Firstly, the Variational Autoencoder (VAE) method is used to estimate and supplement the missing values of e-commerce sales from 2003 to 2012. Then, the mutual information is utilized to assess the rationality of the five predictive indicators selected, while the B-spline-based kernel density regression estimation method is employed to estimate the additive model. Finally, the GM (1, 1) model is used to predict the values of the five predictive indicators for the next three years, which are then incorporated into the established additive model to obtain the forecasted e-commerce sales for the next three years.展开更多
文摘准确预测电子商务销售额对电子商务业高质量发展具有重大意义。本文以贵州省为例,基于收集到的5个预测指标,建立可加模型对贵州省电子商务销售额未来三年进行预测。首先使用变分自编码器(VAE)方法对2003~2012年电子商务销售额缺失值进行估计补充,然后使用互信息评估5个预测指标被选择的合理性,同时使用基于B样条的核密度回归估计方法对可加模型进行估计。最后使用GM (1, 1)模型预测未来三年5个预测指标值,然后带入到已建立的可加模型中,从而得到电子商务销售额未来三年的预测值。Accurate prediction of e-commerce sales is of significant importance for the high-quality development of the e-commerce industry. Taking Guizhou Province as an example, this paper establishes an additive model to forecast the e-commerce sales in Guizhou Province for the next three years based on five collected predictive indicators. Firstly, the Variational Autoencoder (VAE) method is used to estimate and supplement the missing values of e-commerce sales from 2003 to 2012. Then, the mutual information is utilized to assess the rationality of the five predictive indicators selected, while the B-spline-based kernel density regression estimation method is employed to estimate the additive model. Finally, the GM (1, 1) model is used to predict the values of the five predictive indicators for the next three years, which are then incorporated into the established additive model to obtain the forecasted e-commerce sales for the next three years.
