EF can be described as (Hits_s/Ns)/(Hits_t/Nt), where Hits_s is the #of actives in selected front fraction of ranked list, Hits_t is the quantities of actives in database, Ns is the total # of compounds in the selected fraction of the database, and Nt is the quantities of compounds in the database. You can plot EF curve vs database %; and decide if your model is suitable for screening potential actives from a large database.
In simple terms, if Ndiv is the number of actives in a random subset of a large chemical collection and Nvs is the number of actives in a subset of the same size designated using a rational approach, such as virtual screening, then Nvs/ Ndiv is the enrichment rate. It tells you by how much your virtual screening algorithm is better than a random number generator. Note that Ndiv is also the hit rate in an experimental high-throughput diversity screen.