Acollaborative Filtering Recommendation Algorithm Based On User Ratings And Genetic Algorithm

With the rapid development of the Internet, people's lives have changed dramatically, but how to find their own needs from a huge amount of information has become increasingly difficult. In this context, recommendation system came into being,and play a tremendous role;recommendation system in reduce the information overload problem there are many Web brought about many negative influences play more and more important role, and on these sites, users are often very likely through voting score to express their preferences for a range of goods or services.Collaborative filtering recommendation algorithm is widely used at present. It analyzes the user interest, to find the similar user of specified user in the user group, these similar user evaluation of certain information, form the system prediction of preference information about this from the user specified in the user group. Commonly used similarity calculation method with cosine similarity, Pearson correlation coefficient method, but these similarity computation methods are usually more complicated formula, thus leading to recommended in the process of similarity calculation taking too much time, reduced the efficiency of recommendation. A new similarity calculation method based on genetic algorithm and user rating information is proposed in this paper.Firstly, a vectorpx,y=(px,y(0),px,y(1),...,Px,y(C-c)is proposed,The number of elements is C-c+1(for example, C=5, c=1, element number is 5).Px,y(i)=a/bindicates that two user x, y on the same item score difference is i times a and at the same time are the two user rating items over the number of the ratio of b.Secondly, a weight vectorq=(q(0),...,q(C-c) is proposed,The value of each element q(i)is between [-1,1].Each element q(i) is used to measure the importance of px,y(i) for computing the similarity between the two users.A new similarity calculation method is composed of these two vectors. The best weight vector is obtained by genetic algorithm.Finally, The above new similarity calculation method is carried out in FilmAffinity and Movielens data sets. Through the training set obtained recommendation model, then utilizes individual q in the population to the training set to predict recommended, get the individual q corresponds to the MAE systems if less than a given threshold, then the individual is the best individual, will apply to the test set performance testing. Compared with the traditional methods, this method has certain improvement in the performance of the system, and the efficiency of the recommendation is also improved.