The unreasonable roasting temperature will lead to the decrease of pellet quality, the over-roasting has little influence on the compressive strength of pellets, but it increases the heat loss and the energy consumption. While under-roasting is unfavorable to reduction degree and reduction swellability of the pellets. In view of the fact that the roast temperature control system cannot give real-time feedback and that there are many factors affecting the roasting temperature, and in this paper, the data mining and regression analysis of the roasting temperature are carried out through correlation analysis of SPSS and the regression analysis of MATLAB. Finally, it is determined that the hood temperature and windbox temperature have great effect on roasting temperature, which is suitable for design of roasting temperature optimization and control system to ensure the stable roasting temperature and improve pellet quality.