A Recommendation Model Based On Attention Network With Social Information

Due to the continuous rise of the Internet and the continuous development of science and technology,the output of information shows an explosive situation,and this situation has led to information overload.At one time,the main means for people to acquire knowledge was to search according to different classifications,but it has since evolved to actively search through search tools(for example,search engines).But the above methods are based on the information that consumers know their own needs as well as having the desire to explore knowledge according to their needs.When the desire to actively explore reaches a bottleneck,the problem of information overload will be particularly prominent.At this time,a new type of information acquisition and output method that can explore the user's needs and satisfy the user's interest will come into being.This is the recommendation system.The main task of the recommendation system is to communicate information producers and information consumers.On the one hand,the information produced by information producers is pushed to the target customers;on the other hand,information consumers can find information of interest.This process does not need to explicitly provide target demand.The deep learning method has achieved remarkable results in many fields.Using the deep learning method to solve the recommendation problem is one of the hot research problems of the current recommendation model.The neural collaborative filtering(NCF)recommendation framework is a kind of deep learning representative method that uses neural network to extend traditional collaborative filtering.However,both the traditional recommendation model and the recommendation model based on deep learning will face the problem of data sparsity.One way to alleviate data sparse problems is to use the side-information such as social relation.The current neural collaborative filtering method fails to introduce social information to improve the recommendation effect.In order to improve the quality of recommendation,this paper uses attention network to expand the definition of user and item latent vectors while integrating social relation data,so as to build an attention recommendation model with social information.The main research contents are as follows:(1)Aiming at the common problems of data sparsity and cold start in recommendation system,social relation data is introduced to improve recommendation quality.It is a common solution to alleviate the data sparsity problem by using side-information.Here,social information is used to expand the data content,which is essentially to reduce the data sparsity through auxiliary information,so that the recommendation model can obtain richer content.(2)The attention network in deep learning is used to model the latent vectors containing social information.In order to enable the deep learning-based recommendation model to make effective use of social information,the definition of user and item latent vectors need to be extended.A reasonable solution to effectively integrate social information and extend the latent vectors is to use the attention mechanism network.(3)Based on the neural collaborative filtering framework,the final recommendation model is obtained through the extension of horizontal and vertical associations.The neural collaborative filtering framework has been proved to be an effective framework for the recommendation tasks.Based on this framework,horizontal and vertical association extension can be carried out to obtain new recommendation model instances more conveniently.Finally,a series of experiments are carried out to verify the effectiveness of the proposed model,the necessity of the overall construction of the model and the influence of model parameters.At the same time,the experiment also shows that the model proposed in this paper has the effect of alleviating cold start.