Graph meassures on the fee graph
I love this work and the start. I had a quickl look at your source code and it seems that you are looking at the friendship / neighbor graph of payment channels and compute your measures for suggesting nodes / channels.
as explained in this video https://youtu.be/L39IvFqTZk8 if you want to optimize your routing fees it might be interesting to look instead at the fee graph. This is a little bit tricky as the fee consists of a basefee and a dynanamic feerate that depends on the amount you pay. I myself am not sure yet how to model this feemodel in a good way to a fee graph. in my video I assumed a constant feerate and was just optimizing for the feerate..
Also you could have a look at lib_autopilot.py: https://github.com/lightningd/plugins/tree/master/autopilot