A Search-based Method for Forecasting Ad Impression in Contextual Advertising is a paper from Yahoo to forecast ads monetization based on previous exposition of the similar ads. The estimation is precise when there is a large amout of previous data available.
This is a seminal paper, but there is a lot of room for improvement. I assume that there are three types of ads:
- Those always shown such as "real estate"
- Those with periodic peak and a slow growth like "u2 concert"
- Those with instananeous and fast growth like "swine flu vaccine"
It seems that the paper is not addressing type 3
No comments:
Post a Comment