[discussion]Is there any strategy to sample the github actors for Github bot detection?

[bifenglin]

I have loaded the top 500, 000 active repos and checked out the participating actors. But there is still lots of data, so we have to choose a strategy to sample the actors which could be a gound truth data for Github bot detection.

type	count
Bot	1,112
Organization	250,183
User	9509701

[bifenglin]

We take the strategy that order the count of event and choose the top 10000 actors. Since it has enough event data to figure out whether bot or user.