Study and Application of Mineral Products Sales Forecasting Models Based on Market

Abstract

Based on the sales data of a mining enterprise and historical socioeconomic data during 2004-2010, this paper presents a comprehensive study on fluctuation pattern and influencing factors of the sales data, aiming at constructing sales forecasting models accord with mining enterprises' characteristics in product sale time series, namely, multi-dimensionality, nonlinearity, and greater amplitude, etc. Three different forecasting models are investigated for short term sales price prediction and sales volume prediction of a mining enterprise. The adopted forecasting models are time series resolution, auto regression-moving average (ARIMA) and artificial neural network (ANN). They are compared in performance. Based on above-mentioned models, this paper establishes a forecasting model for mineral product sales by incorporating BI techniques with combination forecasting techniques. In addition, the models have been applied to forecasting sales. The result shows that combination forecasting models is capable of inhibiting various types of noises and singular points of product sales series with relatively high accuracy. Moreover, the forecasting accuracy of combination models is better than that of any individual forecasting model. The former can provide effective support to mining enterprises in decision-making and sales management.