Click-through Rate Prediction In Practice: A study of a click-through rate prediction system
Ståhlberg, Kurt-Eerik Eerikinpoika (2019)
Ståhlberg, Kurt-Eerik Eerikinpoika
2019
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2019060314254
https://urn.fi/URN:NBN:fi:amk-2019060314254
Tiivistelmä
Digital advertising is a huge business with tough competition. One of the ways to be more effective in the business is to serve better chosen ads to each user. One way to improve the ad selection is to predict the click-through-rate of each prospective ad and then select the one with the highest predicted CTR. In this thesis three possible choices – naïve, linear and factorization machine – for a predictive model are studied, tuned, evaluated and their results are compared with factorization machine showing the best results.