Research On Recommendation Algorithm Based On Deep Leaking

With the increase in the number and variety of goods in e-commerce sites,consumers need to spend more time on the selection of goods,such as information overload phenomenon plaguing people selection of effective information.In order to solve the above problems,recommendation technology enters people's lives,collaborative filtering recommendation is the most widely used recommended technology.However,with the growing number of people in the network,the data in the collaborative filtering recommendation system becomes more and more sparse,which will affect the accuracy of similarity between users,and it will make it difficult to cluster a large number of users into different groups,which affect the choice of the nearest neighbors.And the quality of the similar groups has a big influence on the effect of the recommendation system.Facing above situation,this paper studies how to improve the recommendation quality of the recommendation system from the aspects of how to improve the clustering,alleviate the data sparseness problem and calculate the user similarity.First of all,in order to solve the problem that limitations of K-means rely on distance clustering and lead to similar users can not gather in the same group and affect the quality of recommendation,in this paper a collaborative filtering recommendation algorithm based on deep belief networks was presented.The method builds the user's feature matrix according to the user and the item rating matrix,and then uses deep belief nets to train the user characteristics,and then perform the user clustering operation according to the obtained top neuron probability and calculate the correlation degree between the target user and the user in the group,and finally selected the users with high similarities as the neighbor users to recommend for the target user.Secondly,aiming at the problem that the traditional collaborative filtering recommendation algorithm can not solve the problem of data sparseness and the shortcoming of similarity computation well,a recommendation algorithm of deep belief networks fusing grade deviation was presented.The method uses the deep belief networks to conduct the clustering operation and calculated the similarities of the users in the same group by grade deviation,and finally selected the users of high similarities with the target users for some relevant recommendations.Finally,experiment was carried out to verify the proposed algorithm,and compared with other related algorithms to analyze the experiment results.