The invention belongs to the field of machine tool thermal error measurement and specifically discloses a numerical control machine tool spindle thermal deformation prediction method and system. A modeling module is used for establishing a neural network model. The neural network model adopts motion state data and thermal deformation state data of a machine tool before the current moment as the input and adopts the spindle thermal deformation amount within a period of time after the current moment as the output. A data acquisition module is used for collecting the machine tool spindle current,the spindle speed, the environmental temperature and the spindle thermal deformation amount to establish a training set. A model training module is used for inputting data in the training set into the neural network model to implement training. A thermal deformation prediction module is used for acquiring the motion state and thermal deformation state data of a machine tool to be predicted in real time and inputting the motion state and thermal deformation state data to the trained neural network model to realize prediction on the spindle thermal deformation. Through the numerical control machine tool spindle thermal deformation prediction method and system, the prediction effect is good, and the influence of thermal errors of the machine tool in the machining process can be effectively reduced. The numerical control machine tool spindle thermal deformation prediction method and system are suitable for prediction on thermal deformation of a machine tool spindle without a built-in sensor.