http://lattes.cnpq.br/2713418370130387; SILVA, Raissa Matias da.
Resumo:
One of the key challenges in constructing a Bayesian network (BN) is defining the node probability tables (NPT). For large-scale BN, learning NPT through domain experts knowledge elicitation is unfeasible. Previous works proposed solutions to this problem using the concept of ranked nodes; however, they have limited modeling capabilities or rely on BN experts to apply them, reducing their applicability. In this work, we propose and evaluate three methods to solve the problem. First, an expert system based on production rules. Second, a method using a brute-force algorithm to identify a set of possible combination. Finally, a method using genetic algorithm to define NPTs with no ranked nodes-specific knowledge. To validate this approach, it was executed an experiment with a BN already published in the literature. Results demonstrated the advantages and disadvantages of each method depending on time, memory availability and parents node quantity. By using one of the presented solution, a practitioner can accurately define NPTs without understanding the concept ofranked nodes.