Reference: Byun, J.-E. and Song, J. (2021). A general framework of Bayesian network for system reliability analysis using junction tree. Reliability Engineering & System Safety, 216, 107952.
As for the transport example we have now, can we apply the same method to these two larger networks?: (1) Sioux Falls (SF) benchmark network and (2) EMA (Eastern Massachusetts) highway network?
For SF, there are three destination nodes 9 (airport), 14 (hospital 1) and 19 (hospital 2). Origin nodes are 1, 2, 3, 6, 7, 10, 12, 13, 18 and 20 that represent major intersections. This creates 3 10 = 30 OD pairs.
For EMA, there is one destination node 1 (representing the city centre). Origin nodes are all other 73 nodes, leading to 1 73 = 73 OD pairs.
I do not think we need to verify the result since we have already proved correctness of the functions by the toy method.
A more important issue is if the computational complexity (i.e. the number of CPM rows to be created) would be affordable.
Could we test this by using the functions developed so far? My hope is that the computational cost will remain affordable by limiting the number of shortest routes (by 10, for example).
As for the transport example we have now, can we apply the same method to these two larger networks?: (1) Sioux Falls (SF) benchmark network and (2) EMA (Eastern Massachusetts) highway network?
For SF, there are three destination nodes 9 (airport), 14 (hospital 1) and 19 (hospital 2). Origin nodes are 1, 2, 3, 6, 7, 10, 12, 13, 18 and 20 that represent major intersections. This creates 3 10 = 30 OD pairs. For EMA, there is one destination node 1 (representing the city centre). Origin nodes are all other 73 nodes, leading to 1 73 = 73 OD pairs.
I attach topology data files of the networks. EMA_net.txt ema_node_km.txt SiouxFalls_net.txt SiouxFalls_node_inch.txt
I do not think we need to verify the result since we have already proved correctness of the functions by the toy method. A more important issue is if the computational complexity (i.e. the number of CPM rows to be created) would be affordable. Could we test this by using the functions developed so far? My hope is that the computational cost will remain affordable by limiting the number of shortest routes (by 10, for example).