The thermal elongation of the headstock of the planer type machines will affect the processing accuracy of the workpiece. The thermal stiffness mechanism of spindle box was analyzed, and the relationship model between temperature rise and displacement was established by using RBF neural network. By measuring the key point temperature of the headstock, the temperature change of the point at the next moment was predicted, then the changing regularity of the thermal stiffness of the headstock was obtained. The validity of this method is verified by comparing the predicted values and measured values of thermal stiffness.