CTR prediction based on click statistic

A new click through rate predicting formula which is based on statistics of clicks and shows is described. It is a result of increasing of log-likelihood given a constraint on formula’s variance at small banner history. Form of this predicting formula is derived in different approaches. High risk problem of formula at small banners history is discussed.
Authors
Research areas
Published in
NIPS Workshop: Machine Learning in Online Advertising
Date
28 Dec 2010