Decomposition-based Multi-objective Backtracking Search Algorithm for Personalized Recommendation

In this paper, a decomposition-based multi-objective backtracking search algorithm (DMOBSA) is proposed for personalized recommendation (PR) problems. In the proposed DMOBSA-PR, an effective individual representation is designed based on the characteristics of PR problems. Moreover, the new crossover and mutation operation in the framework of the BSA algorithm are introduced to update individuals. Furthermore, a decomposition-based multi-objective framework is utilized to balance accuracy and diversity metrics. Finally, the simulation results on Movielens 100K dataset show that DMOBSA-PR is effective compared with other algorithms.