A Community Division-Based Evolutionary Algorithm for Large-Scale Multi-Objective Recommendations

Multi-objective evolutionary algorithms (MOEAs) have been demonstrated to be competitive in recommender systems. In most of the exiting MOEA-based recommendation algorithms, individuals in the population are usually represented in the form of matrix encoding, where rows represent users and columns represent recommended items. However, as the number of users and items increases, the search space of these MOEA-based methods increases exponentially. To tackle the issue, in this paper, we suggest a community division-based evolutionary algorithm named ComEA for large-scale multi-objective recommendations, where a community division-based reduction scheme is proposed to greatly reduce the search space of MOEA from both row and column perspective. To be specific, a user network is firstly constructed by using the user-rating information. Based on the constructed network, we propose a community division-based row reduction strategy to reduce the dimensionality of rows, and then propose a community division-based column reduction strategy to reduce the dimensionality of columns. With the proposed community division-based reduction strategies, the search space of the proposed ComEA can be greatly reduced. Experimental results on
movielens, food
and
douban
datasets demonstrate the superiority of the proposed algorithm over several state-of-the-art multi-objective algorithms for large-scale recommendation problem.