Grammatical Evolution in a Matrix Factorization Recommender System

This paper presents preliminary results of using grammatical evolution to evolve expressions that calculate the user/item features used in the matrix factorization recommendation algorithm. The experiment was performed primarily to determine whether grammatical evolution can be applied to this field, and to compare the results with those of the 'traditional' algorithm. For the purpose of the experiment, we used the CoMoDa dataset, which features realistic data collected over five years. The preliminary results are promising and offer a lot of possible future work, some of which is discussed at the end of the paper.