To evaluate an exploration system, the prize-penalty function,
GmΣk=1 ψkNk{1+γk nkΣi=1 nkΣj=1 jkijφkij}
has been proposed, where G is a gain in the system, (Ψ
k), (N
k), (γ
k), (η
kij), and (φ
kij) are matrixes of prize, sample number, penalty coefficient, identification probability and penalty, respectively, m is number of kinds of information, and n
k is number of groups in the kind of information denoted by k. The sufixes i and j indicate that the sample belonging to the group i is seemed to belong to the group j. The diagonal elements of matrixes (η
ij)k and (φ
ij)k correspond to correct identification, while the other elements to wrong identification. If the matrix (η
ij)k is diagonal, then the iden-tification is perfect. All of the diagonal elements of the matrix (φ
ij)k are zero, while others are negative-The larger absolute value of a element of the matrix corresponds to the more serious mistake. The values of prize, penalty coefficeint and penalty matrixes are given in accordance with the purpose of the exploration, which is represented especially by the penalty matrix.
View full abstract