In this paper, the decision table is used as a tool for representation
of the test knowledge. With that table, the algorithm for building op
timal decision trees, which embody the solution for the test sequencin
g and diagnosis problem, is analyzed. Some improvements in the algorit
hm are also proposed for better efficiency. Furthermore, in order to d
escribe more complicated situation, the conditional probabilities are
included in the decision table, called as conditional decision table.
Different approaches for generating optimal conditional decision trees
, based on the information theory and entropy, are proposed. Finally,
the algorithm for building dynamic procedures is also presented in thi
s paper. Copyright (C) 1996 Elsevier Science Ltd