Recommender Systems (RS) have proven to be a successful personalization technique in this era of ever increasing information overload. Among many available recommendation techniques, Collaborative Filtering (CF) is the most popularly used. However, most of the CF applications use single ratings for recommending items and the use of multi-criteria ratings in the recommendation process is still under-explored. This paper proposes multi-criteria RS based on Adaptive Genetic Algorithm (AGA). The AGA design, which updates the crossover and mutation rates dynamically, is employed to model the users' preferences for multi-criteria ratings on different attributes of items. The AGA optimizes a user's preferences for different attributes in the form of a weight vector. Thus, the AGA finds an individual optimal weight vector in relation to each user. The weight vector is used to recommend items to the respective user. The experiments are conducted on Yahoo movies, a well-known multi-criteria rating dataset. The experimental results confirm that the AGA based multi-criteria RS outperforms the traditional single criteria based Collaborative Filtering RS and the simple GA based multi-criteria RS.