Aiming at the problem that the prediction value lags behind the actual compensation value in the CNC machine tool thermal deformation error compensation, a thermal error advance prediction solution based on the long and short-term memory (LSTM) neural network algorithm was proposed, and the analytical process of LSTM neural network algorithm was discussed in detail. The thermal error advance prediction model of the neural network algorithm based on LSTM was established, and the key temperature measurement points and the thermal deformation error advance prediction experiment were carried out. The experimental results show that the maximum residual, the average value and the mean square error of the actual value of the thermal deformation error and the predicted value are all within the controllable range, and the accuracy of the advance prediction is 86.3%, which further demonstrates the effectiveness of the advance prediction of the thermal deformation error of the machine tool.