A Personalized Recommendation Algorithm Based on MOEA-ProbS

As a technology based on statistics and knowledge discovery, recommendation system can automatically provide appropriate recommendations to users, which is considered as a very effective tool for reducing information load. The accuracy and diversity of recommendation are important objectives of evaluating an algorithm. In order to improve the diversity of recommendation, a personalized recommendation algorithm Multi-Objective Evolutionary Algorithm with Probabilistic-spreading and Genetic Mutation Adaptation (MOEA-PGMA) based on Personalized Recommendation based on Multi-Objective Evolutionary Optimization (MOEA-ProbS) is proposed in this paper. Low-grade and unpurchased items are preprocessed before predicting the scores to avoid recommending low-grade items to users and improve recommendation accuracy. By introducing adaptive mutation, the better individuals will survive in the evolution with a smaller mutation rate, and worse individuals will eliminate. The experimental results show that MOEA-PMGA has a higher population search ability compared to MOEA-ProbS, and has improved the accuracy and diversity on the optimal solution set.