The use of Artificial Neural Networks for extending road traffic monitoring data spatially: An application to the neighbourhoods of Benevento

In this paper, the use of Artificial Neural Networks (ANNs) for spatially extending road traffic monitoring data is studied. The problem consists of estimating the traffic flows on some links of an urban road network knowing the corresponding data on some other links of the network (monitored links). In a previous paper, the authors studied this problem referring to a whole city obtaining promising results. Starting from these results, here we test if to limit the number of monitored links and non-monitored links to a neighbourhood of a city improves or not the results. These results are useful in medium and large cities where other parts do not at all influence some parts of the network and each neighbourhood can be studied independently from the others. To obtain these results, we have partitioned the network of Benevento in six neighbourhoods and trained six different ANNs with simulated data. Numerical results show that to limit the area analysis improves the results significantly with respect to consider the whole network.