Constructing grey prediction models using grey relational analysis and neural networks for magnesium material demand forecasting

In terms of environmental protection, magnesium is a lightweight material that has been widely used to manufacture components for electronics. By forecasting the demand for magnesium materials, we can evaluate its prospects in the related industries. Grey prediction is appropriate for this study, because there is limited available data on the demand for magnesium, and it does not coincide with the statistical assumptions. Therefore, this study applies the GM(1,1) model, which is the most frequently used grey prediction model, to forecast the demand for magnesium materials. To improve the accuracy of predictions with the GM(1,1) model, its residual modification was established by the neural network. In particular, this study used grey relational analysis to estimate the weight of each sample that was required to avoid unreasonably treating each sample with equal importance in the traditional grey prediction. The forecasting ability of the proposed grey residual modification models was verified using real data regarding the demand for magnesium materials. The results showed that the proposed prediction model performed well compared with the other prediction models considered. •A new grey prediction model is proposed for magnesium material demand forecasting.•Applying GRA to estimate the importance of each sample towards a generating sequence.•A simple residual modification mechanism was set up without building a residual model.•The proposed model performed well compared with the other grey prediction models.