Approximate reliability expressions using a decision tree approach

In this paper a decision tree (DT) algorithm, belonging to the family of rule generation methods, is employed to obtain an approximate reliability expression (ARE) of a network. The main idea is to employ a classification technique, trained on a restricted subset of data, to produce an estimate of the reliability expression (RE), which provides reasonably accurate values of the reliability. The algorithm develops a tree by recursively dividing a collection of random examples (training set) for the network at hand, on the basis of the state (operating or failed) of network components. This produces a classifier that can be easily transformed into a set of intelligible disjoint rules. In the examples presented the AREs, built from a very small fraction of the total state space, produce very close reliability estimates, with errors less than 2 %. The excellent results obtained in the experiments show the potential of the method for the evaluation of the reliability of a system through an ARE.