GUARD: A Genetic Unified Approach for Recommendation

Abstract

Recommender systems suggest new items to users based on his/her interests. They appear in distinct contexts, such as movies, e-commerce, search engines, program guides for digital TV. In this work we propose a framework to generate ranking functions for items recommendation based on genetic programming, which we call GUARD (Genetic Unified Approach for Recommendation). The framework was developed under a collaborative filtering approach, but can be easily extended to work with content-based or hybrid recommender systems. The GP-based framework is flexible, works under a multi-objective approach and is able to combine different features and deal with data uncertainty and noisy. GUARD generates ranking functions considering four different measures: precision, recall, diversity and novelty. The framework was tested in the scenario of movies recommendation, using the Movielens 100k and Movielens 1M datasets. The results obtained were compared to those generated by PureSVD, the state-of-the-art algorithm for collaborative filtering. Considering the Movielens 100k, the results of precision and recall were superior to those of SVD. The framework can also generate more diverse and novel recommendations, with a small loss in precision. For Movielens 1M, the results are not as good as those obtained by PureSVD, but the functions slighly sacrifice accuracy over simplicity.