Politecnico di Torino (logo)

Click-Through Rate Prediction

Klodiana Cika

Click-Through Rate Prediction.

Rel. Paolo Garza. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Informatica (Computer Engineering), 2021

PDF (Tesi_di_laurea) - Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives.

Download (1MB) | Preview

Nowadays the internet has drastically changed the advertising industry, and it continues to change as new technology and platforms are released. The success of any advertisement campaign lies in reaching the right class of target audience and eventually convert them as potential customers in the future. In the last few years we can notice a considerable growth of the market of online advertising, which becomes very important, and provides a major source of advertising revenue. In order to measure a campaign effectiveness there exist a variety of metrics and one of them is Click-Through Rate (CTR): a ratio showing how often people who see an advertisement end up clicking it. CTR prediction means capturing user’s dynamic and evolving interests from their behavior sequence and answering to the question: How likely is the user to click on the advertisement? For performing prediction different techniques of Machine Learning are exploited. Machine Learning is a sub field of Artificial Intelligence.While AI is the broad science of mimicking human abilities, machine learning is a specific subset of that trains a machine how to learn. The first approach to Machine Learning theory was done in the ’60s for the American defense industry, in order to understand if the machine can be capable of independent reasoning. This research presents a CTR prediction system that analyzes several factors to predict if an advertisement will receive a click or not. The analyzed dateset in the paper is the Avazu dateset that collects 11 days of user behavior data in order to build and test prediction models. Firstly a detailed analysis is performed in order to better understand the user and extract the important features. Secondly 9 a dashboard is created for the purpose of visualizing the numerous data in a more efficient way. Moreover before proceeding with building the models and the training phase, a pre-processing step is needed with the aim of preparing the data. This work is done in collaboration with Machine Learning Reply which is a pioneering company in the field of Machine Learning and Artificial Intelligence.

Relators: Paolo Garza
Academic year: 2020/21
Publication type: Electronic
Number of Pages: 83
Corso di laurea: Corso di laurea magistrale in Ingegneria Informatica (Computer Engineering)
Classe di laurea: New organization > Master science > LM-32 - COMPUTER SYSTEMS ENGINEERING
Aziende collaboratrici: Reply Consulting Srl
URI: http://webthesis.biblio.polito.it/id/eprint/19217
Modify record (reserved for operators) Modify record (reserved for operators)