An Experience-Feedback Algorithm of D-S Evidence Theory
Baowen Hu, Bo Shen, Qing Liu · 2013
D-S evidence theory is a kind of important method of data fusion to get accurate prediction.In this paper, we propose an improved method in which we build an experience-feedback mechanism for reasoning process.Then prediction accuracy is fed backed to a new round of fusion process in the form of weights to improve new fusion results.We introduce two feedback algorithms and conduct an analysis through comparing some examples.Further, to solve the problem of cold start, we also suggest a method with the generation of random numbers.Simulation results show that the proposed algorithms can not only improve the performance of data fusion and the accuracy of forecast effectively, but also solve the problem of evidence conflict in D-S evidence theory.The information fusion technology is a kind of information process in order to make proper decision and credible predication through automatic analysis and optical synthesis of relevant observation data provided from various sensors utilizing computer technology.One of the main methods for information fusion is D-S evidential theory.The theory of evidence can fuse information provided by multiple sensors, thus reducing the uncertainty of the information.