Probabilistic system i.e. Bayesian system, learning method for determining clinical therapy procedure of patient, involves changing probability distributions depending on quality criteria such that probabilities of edges are increased

摘要

The method involves representing a probabilistic system i.e. Bayesian system, by a graph (AG), and assigning variables to nodes in the graph based on a data record (D) of a patient. Graphs (G1-G4) with edges (K1-K6) between the nodes are generated with edge probability distributions (PD, PD'). A parameter e.g. therapy parameter, is learned for a part of the graphs, and quality criteria (SC) are determined for the learned graphs. The edge probability distributions are changed depending on the determined quality criteria such that probabilities of the edges in the graphs are increased. An independent claim is also included for a computer program product comprising a program code for executing a method for learning a probabilistic system.