In the field of programmatic advetising based on the ecosystem called realtime bidding, it is important to know, how successful an ad impression will be. Click-through rate prediction is one of the biggest challenges in online advertising. In this thesis we use factorization machines to predict the clickthrough rate based on data about the ad, website, user etc. We describe the process of data preparation, feature selection, implementation and testing. The goal is to improve the current solution in company Zemanta d.o.o. which is based on logistic regression. With local testing and online A/B testing we reach our goal and contribute to improving the service and financial performance of the company
|