Using PCA-BP Neural Network Algorithm to Forecast the Cost of Rare Earth Products

Abstract

The BP neural network pattern and algorithm are widely used in the manufacturing industry. The article develops the BP neural network pattern into the resource industry.Considering the multicollinearity of the influencing factors, this article constructs a combination model of principal component analysis and BP neural network (PCA-BP), and makes use of the excellent characteristics of this model to predict the price of rare earth products.In this text, monthly dysprosium oxide price data and 33 influencing factors are selected for model verification and application, and it is found that the prediction results of the combined model are superior to the traditional BP neural network pattern in terms of emulation capability, mistake level, and convergence accuracy. The results show that PCA-BP combined model combines the advantages of factor analysis method and neural network, effectively improves the prediction accuracy of neural network, and can accurately predict the change trend of dysprosium oxide price.The combination model has broad application prospects in resource industry price prediction.