Knowledge Acquisition in PROSTANET - A Bayesian Network for Diagnosing Prostate Cancer

The manual development of a Bayesian Network (BN) is considered more an art than a technique, which mostly depends on the interaction between knowledge engineers and human experts. As a conse- quence, an explanation facility constitutes a very useful tool. The current paper focuses on the process of building Prostanet ,aB N designed to help to diagnose prostate cancer. In this case, the explanation capabili- ties provided by Elvira(2) have helped, not only to define the structure of the network, but also to get and refine its probabilities, which have been estimated subjectively by an urologist.