Click-through rate (CTR) is one of the most important measurements in online advertising that tells us how successful a certain ad is. In this thesis we work on CTR prediction of ads on individual websites for newly created ads that have not been published on any website in the past. We also work on evaluation of attributes that describe these ads. We use data about ads and ad clicks from Celtra d.o.o. and process it in different ways. Then we apply different machine learning methods - random forest, k-nearest neighbors, and matrix factorization. For attribute evaluation we use RReliefF and the difference of variance. We find out that the website, on which an ad is published, and the size of an ad influence CTR the most. Based on our predictions of CTR, a company can decide on which websites they should publish the ad thus enlarge the number of clicks on the ad.
|
