The invention relates to a machine tool linear axis thermal temperature sensitive point optimization method based on an independent variable selection criterion, which is characterized in that in different temperature states, a machine tool linear axis positioning error is measured, the independent variable selection criterion is utilized to select an optimal regression subset so as to select an optimal temperature sensitive point, and a thermal positioning error model is established. The method comprises the following steps of: acquiring a temperature sensitive point of a machine tool, acquiring a geometric positioning error model, superposing with the geometric positioning error model to obtain a comprehensive positioning error model, measuring the temperature at the temperature sensitive point of the machine tool in real time, and the like, and effectively solving the problem of collinearity among temperature variables by adopting a temperature sensitive point selection method of a degree-of-freedom adjustment negative correlation coefficient maximum criterion Ra2 and a corrected AIC criterion. And multiple regression analysis is adopted, a temperature sensitive point selection method and a thermal error modeling method are unified, and the prediction precision and robustness of a thermal error model are improved.