USO DE REDES BAYESIANAS OBTENIDAS MEDIANTE OPTIMIZACIÓN DE ENJAMBRE DE PARTÍCULAS PARA EL DIAGNÓSTICO DE LA HIPERTENSIÓN ARTERIAL

Abstract

In the present work, different Artificial Intelligence techniques are combined to model the diagnosis of hypertensive people. To develop the work a data base of Arterial Hypertension was used, which is result of a preliminary study made in five polyclinics of Santa Clara city, with supposedly healthy individuals. One of the ways to model the relations between variables is using a Bayesian network. The computational cost of the learning of a Bayesian network from data, grows with the number of variables and the number of cases, therefore, the problem of identifying a good heuristic to explore the space of possible networks arises. The evolutionary algorithms are being very valuable methods to find good solutions to concrete problems, that is why the Particle Swarm Optimization (PSO) algorithm is used for the network structure search. An extension to the Weak platform (Waikato for Environment Knowledge Analysis) was done, in which the new algorithm becomes part of the global score metrics implemented in the Bayesnet class. The obtained results show good classification of the Arterial Hypertension with Bayesian networks.