Study of iron concentrate price forecasting models based on data mining

Abstract

Based on the operations data of a Chinese mining enterprise and historical socioeconomic data during 2011–2015, a comprehensive study on the fluctuation pattern and influencing factors of the iron concentrate sales price is carried out, according to characteristics of the iron concentrate sales price time series of the mining enterprise, namely, nonlinearity and multi-dimensionality. Different data mining and big data analysis methods are investigated for the iron concentrate sales price prediction of a mining enterprise. The adopted forecasting models are artificial neural network and auto regression-moving average. A comparison of the forecasting accuracy of different forecasting models is carried out. Based on above-mentioned models, a forecasting model for the iron concentrate sales price by incorporating big data analysis with combination forecasting techniques is established and that is then applied to forecasting the iron concentrate sales price of the mining enterprise. The forecasting result shows that the combination forecasting model is capable of inhibiting various types of noises and singular points of the iron concentrate sales price series with relatively high forecasting accuracy. The forecasting accuracy of the combination model is better than that of any individual forecasting model. The combination forecasting model is effective and feasible. The forecasting results can provide effective support to mining enterprises in decision-making.