Abstract: The spindle thermal error of the precision five-axis machining center is the main factor affecting the machining accuracy of the machine tool. It is an effective means to establish the thermal error model of the machine tool spindle through the deep neural network. However, there are many parameters in the deep neural network, and different parameters will lead to huge differences in the effect of modeling. In order to further improve the accuracy of machine tool thermal error modeling, a prediction and compensation method of machine tool spindle thermal error based on intelligent algorithm optimized deep neural network was proposed. Aiming at the mechanical spindle structure of a five-axis machining center, based on the calculation method of maximum mutual information coefficient, the correlation and hysteresis analysis of temperature-spindle thermal displacement were comprehensively included in the screening of temperature sensitive points, and different speed experiments were set up. At the same time, the cooling stage of the machine tool was added to continuously collect temperature and key thermal deformation error data of the spindle. The ant lion optimizer (ALO) was used to optimize the parameters of long short-term memory (LSTM) neural network, and the ALO-LSTM model was constructed. Finally, this method was used to predict the thermal error of the spindle and the actual processing experiments of NAS samples before and after the thermal error compensation of the spindle. The results show that the spindle thermal error prediction model optimized by the intelligent algorithm has better accuracy than the original model, and the accuracy of the compensated NAS sample meets the test standard. Compared with the accuracy of the NAS sample before compensation, the compensation ratio reaches more than 80%, which has certain reference significance for the thermal error compensation of the precision five-axis machining center.