Recommender Systems

The recommender systems are used by many e-commerces and companies that offer products as a service (i.e., streaming) to stimulate their clients consumption while maximizing their satisfaction. Most approaches result in a black box model that can be used solely for the recommendation purpose. This project aims at finding explicit mathematical expressions modeling the user behavior that can be used to explain every recommendation (to the user and to marketing purposes), recommend a set of unpopular products to the correct group of users and discover the characteristics of the most profitable products to guide the engineering of new products.

Graduate students: Cássia de Souza Carvalho

Related publications:

• Coherent recommendations using biclustering.
• Evaluating the performance of a biclustering algorithm applied to collaborative filtering-a comparative analysis.
• Applying biclustering to perform collaborative filtering.
• Predicting missing values with biclustering: A coherence-based approach
• Multi-objective biclustering: When non-dominated solutions are not enough.
• Query expansion using an immune-inspired biclustering algorithm.