大量信息源条件下的组合规则研究

Research of Fusion Rules on Mass of Information

如何解决组合过程中的冲突是D-S证据理论面临的一个难题.现有的Dempster组合规则及大多数改进规则,如Mur-phy的算术平均规则,Smets的开放世界模型,Yager规则,D.P.规则以及PCR规则等,研究的是证据数目较少的情形.本文对大量信息源条件下,各种组合规则的性能进行了分析和研究,提出组合规则在大量信息源情况下的三个评价标准大众意见、幂等性、收敛性;并给出在大量信息源情况下,依据对结果的要求是否保守来选择不同的组合规则的选择策略.最后提出一种结合平均思想达到收敛性的折中组合规则.

One problem in D-S theory is how to deal with conflicting. The existing rules such as the like modified D-S rule and its improvements, say as Murphy＇s average rule, Smets＇ open world rule, Yager＇s rule, Dubois ＆ Prade＇s rule, PCR rule, work only when the source of information is not too much or is limited, in this paper, the condition of mass of information is analyzed, using the rule presented above to test their capability on new condition. Then three criteria on mass of information condition arc given, which are majority opinion, idempotence, convergence. Also the policy of selecting the fusion rule according to whether the result is conservative is presented. Finally we propose a new rule with averages by which we achieve the convergence.