Predicció de clics amb Tensorflow

Abstract

This project is based on a challenge provided by the website called Kaggle, which involves the creation of an automatic learning system that makes possible a prediction of clicks on online advertising. Assuming this challenge, a program that is able to predict ad clicks has been created, but by using new technologies like Docker, which makes possible a virtualization of a customized Linux environment ideal for programming Tensorflow, a library that enables an automatic, fast and flexible learning. The program’s main algorithm makes an automatic learning using a linear regression model that feeds from the database provided by Kaggle, which is destined for this challenge. This database is composed of a 10 day history of ad clicks and no clicks.