Abstract: The policy decision system for mining project investment involves such main variables as mineral product price, production cost, market demand and risk interest level, of which ore price has the dominant role. It is in non-linear variation and difficult to forecast by using classic time series theory, leading to a difficult realization of the optimization of the policy decision system for mining project investment. Taking uranium resource as example, BP neural network and adaptive neuron-fuzzy inference system are used, in combination with time series technique, to build BPNN and ANFIS time series models of the uranium product respectively. Comparative analysis of the forecasts of the uranium product price is made. The result indicates that ANFIS time series can better forecast the price than BPNN time series does.

Key words: ANFIS, BPNN, Time series, Price forecast