A novel method based on Least Square Support Vector Machine(LS-SVM) and genetic algorithm to select the temperature sensors of a Numerical Control(NC) machine tool was presented.The measurement points in a CNC lathe were grouped based on the thermal mode theory,Then,the genetic algorithm was used to determine the positions of optimum sensors.Finally,a thermal error regression model was established by the LS-SVM and a compensation model for the machine tool was given also.The results show that the novel method combined genetic algorithms and LS-SVM well avoids the correlation of the temperature sensors and ensures the accuracy of the model.In the experiments of the CNC lathe,the mean absolute percentage error of the LS-SVM model is 1.89% in axial direction and 2.04% in radial direction,it also can reduce costs and shorten modeling time for less temperature sensors.