Estimation of parameters based on artificial neural networks and threshold of HIV/AIDS epidemic system in Cuba

In this paper, a method of parameter estimation, based upon Hopfield neural networks, is applied to the identification of a model of the HIV/AIDS epidemic in Cuba. This estimation technique presents a number of features that make it especially suitable for the identification of such epidemic models. In particular, its remarkable ability to estimate time variant parameters allows us to outline the evolution of parameters, thus assessing the efficiency of health policies. A complementary aim is the validation of the model, which stems from basic concepts and reasoning in disease modelling rather than from specific physical laws, thus it is not supported by rigorous mathematical results. Indeed, the experimental results show a significant adjustment of simulated data to actual ones, which provides considerable support to model validity. A sequence of real data, corresponding to the HIV and AIDS affected populations in Cuba from 1986 to 2004, has been compared to both simulations of the epidemic model and used to test the estimation method. A detection threshold, which plays a similar role to the basic reproduction number, has been calculated to complement the information about the current situation of the epidemic and the possible evolution perspectives. The obtained results are plausible and coherent according to health experts.