Dempster-Shafer证据理论(证据推理融合)

证据理论起源于1967年Dempster提出的由多值映射导出的上概率和下概率,之后Shafer进一步将其完善,建立了命题和集合之间的一一对应关系,把命题的不确定性问题转化为集合的不确定性问题,满足比概率论弱的情况,形成了一套关于证据推理的数学理论。

一般的集合为:\Phi = \left \{ \Theta \: 1,\Theta \, 2, \Theta \, 3\right \}

其中元素是两两互斥的。

证据推理融合的集合为:2^{\Phi }=\left \{ \phi \left \{ \Theta 1 \right \}\left \{ \Theta 2 \right \} \left \{ \Theta 3 \right \}\left \{ \Theta 1\Theta 2 \right \}\left \{ \Theta 1\Theta 3 \right \}\left \{ \Theta 2\Theta 3 \right \}\Phi \right \}

这里用m表示概率;

证据推理融合的公式为:m(A)=0,A=\phi

                                         m(A)=\frac{1}{1-k}\sum _{B\cap C=A}m_{1}(B)m_{2}(C),A\neq \phi

其中k=\sum _{B\cap C=\phi }m_{1}(B)m_{2}(C)称为规范化因子。

举例说明:假设中国的卫星观测器m_{c}(\Theta _{1})=0.6 , m_{c}(\Theta _{2})=0.1,m_{c}(\Theta _{3})=0.2,m_{c}(\Phi \)=0.1

欧洲的卫星观测器m_{R}(\Theta _{1})=0.5,m_{R}(\Theta _{2})=0.2,m_{R}(\Theta _{3})=0.2,m_{R}(\Phi )=0.1

现在需要求出它们融合后的值。

m^{'}(\Theta _{1})=m_{c}(\Theta _{1})*m_{R}(\Theta _{1})+m_{c}(\Theta _{1})*m_{R}(\Theta _{2})+m_{c}(\Theta _{1})*m_{R}(\Theta _{3})+m_{c}(\Theta _{1})*m_{R}(\Phi )+m_{R}(\Theta _{1})*m_{c}(\Theta _{2})+m_{R}(\Theta _{1})*m_{c}(\Theta _{3})+m_{R}(\Theta _{1})*m_{c}(\Phi ) =0.6*0.5+0+0+0.6*0.1+0+0+0.5*0.1=0.41

m^{'}(\Theta _{2})=m_{c}(\Theta _{2})*m_{R}(\Theta _{1})+m_{c}(\Theta _{2})*m_{R}(\Theta _{2})+m_{c}(\Theta _{2})*m_{R}(\Theta _{3})+m_{c}(\Theta _{2})*m_{R}(\Phi )+m_{R}(\Theta _{2})*m_{c}(\Theta _{1})+m_{R}(\Theta _{2})*m_{c}(\Theta _{3})+m_{R}(\Theta _{2})*m_{c}(\Phi )=0+0.1*0.2+0+0.1*0.1+0+0+0.2*0.1=0.05

m^{'}(\Theta _{3})=0.08

m^{'}(\Phi )=m_{c}(\Phi )*m_{R}(\Phi )=0.1*0.1=0.01

进行归一化m(\Theta _{1})=\frac{m^{'}(\Theta _{1})}{m^{'}(\Theta _{1})+m^{'}(\Theta _{2})+m^{'}(\Theta _{3})+m^{'}(\Phi )}=\frac{0.41}{0.55}\approx 0.75

m(\Theta _{2})\approx 0.09

m(\Theta _{3})\approx 0.15

m(\Phi )\approx 0.02

此上就是证据推理融合。

你可能感兴趣的:(故障诊断,故障监测,python,matlab)