Cost approximations

• Based on two metrics:

  • The Benefit margin of a cluster C
    • Sum of benefit margin of each subscription in C
    • For a subscription S: ?(p(C)) checkpred(nbpred(S)) - nbpred(p(C)) ?(pred(S)) checkpred(0)
    • Can be approximated using simple statistics
      ?(p(C)) (nbsub(C) - Stat_nbmatch(C))
  • Multi-attribute hash table Benefit
    • | T| - number_of_checked_subscriptions

-

Diapositive précédente

Diapositive suivante

Revenir à la première diapositive

Afficher la version graphique