Research On Personalized Precision Recommendation Algorithm Based On Data Fusion

With the development of Internet technology,people generate a lot of data on the network every day.These data are diverse,and they record the internet behavior habits of each internet user.And the amount of information that users browse every day shows an explosive growth,and users are often not easy to find content they are interested in facing massive amounts of information.In order to cope with this information overload situation,the recommendation system can actively recommend content that the user may be interested in based on the user's historical behavior data.However,the amount of network information is increasing faster and faster,and improving the accuracy of the recommendation system in the context of big data has been the main direction of related research.The traditional recommendation system only relies on the user-item binary relationship to model the user's interest,and the accuracy of recommendations can be lower and lower.Problems such as data sparsity and cold start leading to missing features during the recommendation process make it difficult for the recommendation results to meet user needs.Subsequent introduction of a socialized recommendation system with social information proved to have a certain effect on improving the accuracy of the recommendation,but due to the unreliability of the social information,the recommendation results will be affected by the noise in the social information,which also leads to a reduction in the accuracy of the recommendation.Therefore,the key to the recommendation system research is to improve the feature extraction capability of the recommendation model and improve the accuracy of the personalized recommendation system.In a complex real-world application environment,the recommendation system mainly has the following problems:(1)The cold start of user items leads to the lack of user item characteristics,which prevents the recommendation system from personalizing recommendations.For example,if a new user of an online video website does not generate historical data for online viewing,the recommendation system cannot know the user's preference for watching videos,and cannot make personalized video recommendations.(2)There is an immediate problem with user behavior intentions,which leads to a decrease in the accuracy of personalized recommendations.The user's own interest preferences will change over time,and may also be affected by factors such as festivals,geographic location,and other surrounding environments.These instantaneous variable factors related to user behavior lead to deviations in the prediction of the recommendation system and reduce the accuracy of personalized recommendations.In view of the above problems in the current recommendation system,this paper proposes two novel data fusion recommendation models.Aiming at the problem of missing features caused by cold start,a recommendation model combining implicit and explicit trust is proposed,which is applied to the score prediction task in the recommendation system.Aiming at the problem of immediacy of user behavior,this paper proposes a recommendation model that integrates users' long-term and short-term interests,and is applied to the click-through rate prediction task in the recommendation system.The main research content of this article can be summarized as follows:(1)Aiming at the problem of missing features caused by the cold start of user items,to reduce the negative impact of explicit trust relationship noise in the social recommendation on the recommendation model,a matrix decomposition recommendation model combining implicit and explicit trust relationships is proposed.Use implicit trust relationship and explicit trust relationship to complete user features,alleviate the problem of missing features in the recommendation system,improve the accuracy of recommendations,and use the model to score prediction tasks in the recommendation system.(2)To address the immediacy of user behavior,the recommendation model's ability to perceive the user's behavioral characteristics at different times is increased,and a recommendation model is proposed that integrates the user's long-term and short-term interests.The model is used to reduce the noise generated by the irrelevant behavior in all the historical behaviors of the user to the recommendation result,thereby improving the accuracy of the vectorization of the user's behavior characteristics,and the model is used to perform the click rate prediction task in the recommendation system.(3)For the two recommendation models proposed in this paper,experiments are conducted on the public data sets FilmTrust,Ciao,Epinions,and Amazon Dataset to verify that the recommendation model model that incorporates implicit and explicit trust is for ordinary users and cold-start users.Recommendation effect and verification accuracy of the fusion user long-term and short-term interest recommendation model.The experimental results show that the recommendation model in this paper can effectively alleviate the cold start problem and improve the accuracy of recommendation prediction.