Infrastructure assessment: fuzzy regression with neural networks
P.C. Hoffman, Karen C. Chou · 2002
In modeling a problem where the information is dependent on subjective estimations the fuzzy structure of the problem must be included. One approach in formulating the fuzzy structure is fuzzy regression with neural network analysis. The procedure is to use a neural network model that provides both upper and lower bounds to the data. These bounds define an interval model from which a fuzzy model can be developed. Such an approach has been demonstrated on the problem of quality evaluation of injection mouldings. We propose that fuzzy regression with neural networks would prove most beneficial for the implementations of bridge and pavement management systems. The neural network analysis is more computational intense with six input variables. The development of a fuzzy regression model from the upper and lower bounds neural network models becomes more complex.>