| Summary: | The online advertisement industry handles a large quantity of money and users everyday. This industry is always trying to get more efficient, for example, by enhancing the targeting of online advertising campaigns. This pursuit of efficiency on the world of online advertising turned simpler methods of predic- tion unable to report an accurate number of impressions, used to calculate the value of a publisher's inventory. The introduction of concepts like frequency capping made that very clear. There is now the necessity not only to predict the number of visits, but also to predict when this visits will happen, what the user did before going to that website and who he is. In this document that concept will be approached using Data Mining techniques, such as clas- sification and clustering, in order to generate a future ad request log using only past data. This generated results will be perfect afterwards, to be used on simulators capable of calculate important metrics, for publishers and advertisers, for a set of campaigns.
|
|---|