CTR PREDICTOR

Online advertising systems heavily rely on the ability of the machine learning algorithms to predict the add click accurately and reliably. The click-trough rate problem is challenging. On the one hand, the training data set is enormous (it has millions of observations). On the other hand, algorithms for online advertising in real industrial settings require models with a very large number of coefficients (millions of unique features). Thus, add click prediction requires sophisticated machine learning algorithms to deal with both mass data and large feature spaces.

Technologies & methodologies

Github

GitHub is a web-based Git repository hosting service. It offers all of the distributed revision control and source code management (SCM) functionality of Git as well as adding its own features.

Machine Learning

Usage of FTRL-proximal algorithm: performance evaluation using weighted log-loss. Adapted from the kaggle CTR prediction contest. More info about the algorithm here.