Proposing an evolutionary method based on maximization precision of group recommender systems

In this paper we proposed an evolutionary algorithm to maximize precision of group recommender systems and reducing the online calculations. In this method we try to build a transition matrix that's made by an evolutionary algorithm and then we multiply this transition matrix with user-item rate matrix. By this action we go to a reduced dimension space. The characteristic of this space is that the users that are really more similar, would be closer to each other. Also because the dimension of the user-item matrix has been reduced, the online calculations are hugely reduced and if we have a new user, we can easily multiply his rates on the transition matrix and find out that he has to be in which group. We used the average real rate of group users to an item as a metric to evaluate how much this item is suitable for this specific group. At end we compare this method with other methods that also reduce the dimension on the various datasets. Then we show that our method works better. Finally we have a discussion about weakness and strength of our method.