Collaborative filtering for recommender systems with implicit feedback

Recommending the right products to the customers can significantly increase the sales of an e-commerce, and the presence of huge amounts of transactional data makes data-driven solutions the best choice for the recommender systems in many circumstances. In this work, a general overview of the recommendation task is given, then several data-driven methods are compared on a real world company data. In particular, the effort is centered around implicit feedback, i.e. binary data such as sales, and collaborative filtering, that is the usage of community behavior in the suggestions computation. Finally, different ways to handle cold starts, that are new customers, are discussed and compared.