In the era of information explosion, recommender systems have been widely used to reduce information load nowadays. However, mainly traditional recommendation techniques only paid attention on improving recommendation accuracy without considering additional criteria such as diversity, novelty. Moreover, such traditional recommendation algorithms were also struggled with matthew effect, that is, the gap between the popularity of popular and non-popular items grows. Therefore, a multi-objective recommendation model with extreme individual guided and mutation adaptation based on multi-objective evolutionary algorithms (MOEA-EIMA) is proposed in this paper. It maximizes two conflicting performance metrics termed as precision and novelty. In MOEA-EIMA, the iteration of population is guided by extreme individuals, and the adaptive mutation operator is designed for saving the better individuals. The algorithm is tested in several sparse datasets. The experiment results demonstrate the proposed algorithm can achieve a good trade-off between accuracy and novelty.