This task is less urgent but nonetheless important.
At the moment we have only implemented the hospitals-residents algorithm provided by the matching library. The capacity set for this algorithm is currently one (meaning that there is a one-to-one mapping between the buses). Note that the basic stable marriage algorithm is not suitable here because we expect to have different numbers of buses in each network, and stable marriage would require that these numbers are the same. Hospitals residents is a generalisation of stable marriage which allows for this. We have the following options:
try different capacities in one or both networks, so that multiple buses can be merged together
set a threshold condition on the maximum distance allowed between buses which are merged
construct preference lists using other factors (at the moment the only factor is the distance between buses. We could also consider country code, carrier, and other buses properties)
we could compare what happens if we use k-means instead. I would also like to implement a 'basic' approach for comparison purposes where we simply go through the list of buses in one network, and find the closest bus in the other.
There are many other variants of matching algorithms that we can consider in the future.
This task is less urgent but nonetheless important. At the moment we have only implemented the hospitals-residents algorithm provided by the matching library. The capacity set for this algorithm is currently one (meaning that there is a one-to-one mapping between the buses). Note that the basic stable marriage algorithm is not suitable here because we expect to have different numbers of buses in each network, and stable marriage would require that these numbers are the same. Hospitals residents is a generalisation of stable marriage which allows for this. We have the following options:
There are many other variants of matching algorithms that we can consider in the future.