Abstract: As a key evaluation index for the quality of roasted iron ore, the magnetic tube recovery rate (MTRR) was usually postponed intermittently to get until it has been analyzed chemically in off-line way after each and every roasting process completed in shaft furnace. An intelligent model has been developed to predict the MTRR punctually on the basis of intelligence technology. This model consists of five modules served as for data acquisition pretreatment, decision-making, prediction, online modification and quality evaluation. Introduces the model's framework and main functions of those modules. The model has been applied to the optimal operation and control of iron ore roasting process in shaft furnace and laid down a solid foundation for the comprehensive automatic system in ore processing plant. Its obvious benefits are low maintenance cost, good real time function, and high reliability precision.