Abstract:

Due to the short life of inertial guidance system (IGS) and the small sample obtaining from object test data, the usual support vector machine (SVM) prediction algorithm cannot explain the physical significance, and it does not use subject information. This paper first proposes a new kind of kernel function of LS-SVM based on expert knowledge. Evidence theory measures the influence of predicting factors on predicting object according to expert knowledge. The kernel function takes full advantage of the subjection and objection knowledge, so it has a more reasonable physical significance. Finally, the results of a study case show that the LS-SVM drift error coefficients predicting method and the kernel function of evidence theory are reasonable and feasible.