Prediction of satellite clock errors using LS-SVM optimized by improved artificial fish swarm algorithm

The prediction of the satellite atomic clock errors plays an important role in the work on time and frequency. Aiming at the poor performance of short term prediction of navigation satellite atomic clock errors, a method based on the least square support vector machine (LS-SVM) optimized by improved artificial fish swarm algorithm (IAFSA) is proposed to obtain accurate satellite clock errors. The dynamic parameter adjustment function is introduced to improve performance of artificial fish swarm algorithm. Then it was used to choose the penalty parameter and kernel bandwidth parameter of LS-SVM, which could avoid the man-made blindness during parameters selection of LS-SVM and enhance the efficiency of clock errors prediction. The clock data of four typical GPS satellites are chosen and make comparison and analysis with other three models. The results show that the prediction precision of the proposed method has better prediction performance than the traditional methods, which can afford high precise satellite clock errors prediction for real-time GPS precise point positioning system.