#8 describes individual change classification algos. But we need a higher level abstraction that considers all of the individual metrics in conjunction. Additionally, the interface must be made s.t analysts them selves can define their own change hueristics and add them to the bundle. It must be flexible. We cannot say "this is how you define change in Bitcoin". Perhaps a researcher is interested in only a subset of the individual metrics. The hard part of this ticket is designing the interface for that and figuring out the algorithm for combining each output of a change classification to make a descion about clustering. Perhaps this can be a desc. tree or something smarter: https://github.com/arminsabouri/bitcoin-papers/blob/master/privacy/clustering/resurrecting_address_clustering.pdf
#8 describes individual change classification algos. But we need a higher level abstraction that considers all of the individual metrics in conjunction. Additionally, the interface must be made s.t analysts them selves can define their own change hueristics and add them to the bundle. It must be flexible. We cannot say "this is how you define change in Bitcoin". Perhaps a researcher is interested in only a subset of the individual metrics. The hard part of this ticket is designing the interface for that and figuring out the algorithm for combining each output of a change classification to make a descion about clustering. Perhaps this can be a desc. tree or something smarter: https://github.com/arminsabouri/bitcoin-papers/blob/master/privacy/clustering/resurrecting_address_clustering.pdf