A Bayes Fusion Evaluation Method Based on Representation Point Optimization

Abstract

The invention discloses a Bayes fusion evaluation method based on representative point optimization, comprising the following steps: 1, constructing a classic Bayes point estimation model, calculating the Bayes posterior estimated value of parameters (μ, D); 2, using a clustering algorithm to The prior sample data is divided into components n r class n r ≥4, with the n r The cluster center of the class is the representative point to calculate the posteriori estimated value of the parameter; 3, construct the optimization function, calculate the optimization function value, the optimization function includes the deviation function and the information loss function; 4, screen the best representative point according to the optimization function value, Get Bayesian posterior estimates based on representative points. It has the advantages of reasonably and effectively using prior information, quantifying the influence of the number of representative points on information loss and fusion efficiency, and improving the accuracy of test evaluation.

Description

technical field [0001] The invention mainly relates to the field of precision evaluation of a guidance system and a positioning system, in particular to a Bayes fusion evaluation method based on representative point optimization. Background technique [0002] For the accuracy test evaluation of the guidance system or positioning system, the sample size of each stage of the test is small, and the accuracy of the results obtained by direct evaluation is low. Literature [1] (Tang Xuemei, Zhang Jinhuai, Shao Fengchang, etc. Analysis and Evaluation of Small Samples of Weaponry and Equipment [M]. Beijing: National Defense Industry Press, 2001: 1-3) pointed out that in the process of test evaluation, in order to solve the comparison of samples For few problems, the Bayes method is often used for experimental evaluation. Its advantage is that it can fully integrate the prior information of various heterogeneous experiments, thereby improving the accuracy of test evaluation, but the...

AI Technical Summary

Problems solved by technology

The risk of normalizing the prior information into a test point in the national military standard can be avoided by selecting representative points, but when there is a deviation in the prior data, it may be invalid to directly use the clustering algorithm to estimate the representative points, because When the number of representative points is too large, the prior deviation may have a greater impact on the evaluation results; and when the number of representative points is too small, the information loss is too large

Images