Resumo:
Data modeling is a technique that assists decision making and the resolution of several types of problems in real systems. However, choosing the best technique for modeling a real system is not always an easy task, as each system has its specific characteristics and it is not possible to infer that the best method for a given situation will be the best in other contexts. Due to this need for generalization, the present work aims to present a comparative analysis between the performance of nonlinear regression (NLR) and artificial neural networks (ANNs) methods in modeling data sets generated through a design of experiments (DOE), in order to simulate different scenarios in which the methods will be applied. Thus, the data sets were modeled by each of the methods and their responses were assessed using the coefficient of determination (R²). The results showed that the studied methods present statistically significant differences, with a significance level of 5%. In addition, in this research it was concluded that if the experimenters know a priori the nonlinear model that defines the relationship between the input variables and the response variable, nonlinear regression outperforms artificial neural network in terms of R².