The invention discloses a GPSINS combination navigation method and system utilizing a neural network algorithm to realize compensation and correction. The method comprises steps that a combination navigation model based on a GPS and an inertial navigation module is constructed, a Kalman filter is designed to filter combination navigation data of the combination navigation model, and navigation error data of the inertial navigation module is outputted; in a normal GPS signal state, inertial navigation data is taken as input data of a training sample of a neural network model, output data of the navigation data after filtered by the Kalman filter is taken as output data of a training sample of a neural network module, and the neural network model is trained; in a GPS signal loss state, the trained neural network model is utilized to predict an output error of the inertial navigation module, and the predicted output error is further utilized to compensate and correct inertial navigation. The method is advantaged in that the GPSINS combination navigation model is compensated and corrected through utilizing the neural network algorithm, and the inertial navigation system is assisted by the neural network algorithm to output the accurate navigation data in the GPS signal loss state.