A novel method based on least squares support vector machine and genetic algorithm to select temperature sensors of NC machine tool was presented. The measurement points were grouped based on the thermal mode theory. Then the genetic algorithm was used to determine the optimum sensors’ positions. Thermal error regression model was given by the least squares support vector machine (LS-SVM) finally. The compensation model is made for machine tools. The result shows that the two methods of genetic algorithms and LS-SVM were combined well which not only avoided the correlation of the temperature sensors and ensured the accuracy of the model but also saved the cost and shortened the modeling time.