Extraction of rules from Artificial Neural Network for Dutch Porous Asphalt Concrete Pavement

Among a large number of existing ANN, the multilayer perceptron (MLP) with feedforward (FF) architecture is one of the most widely used structures. They are especially useful as function approximator because they do not require prior knowledge of the input data distribution and they have been shown to be universal approximators. However, despite their high degree of accuracy, these connectionist models are difficult to interpret. For many problems, it is desirable to extract knowledge from trained ANN so that the users can gain a better understanding of the solution. This paper applies REFANN (rule extraction from function approximating neural networks) to analyze the performance of porous asphalt concrete (PAC) pavement. The REFANN rules generated from the data of 72 motorway sections are then compared to the rules generated by regression trees.