Research On Personalized Recommendation Algorithm Based On Hybrid Straight

Recommendation systems are used when searching online databases.As such they are very important tools because they can help users to make a choice between different potential choices and overcome the selection problem caused by information overload.They can be used on e-commerce websites and have attracted considerable attention in the scientific community.To date,many personalized recommendation algorithms have aimed to improve recommendation accuracy from the perspective of vertex similarities.But they tend to ignore the diversity of recommendations.Unilateral improvement of recommendation accuracy will lead to “Information Cocoons” phenomenon,making it difficult for people to expand their knowledge world.Therefore,this paper proposes a personalized recommendation algorithm based on hybrid strategy to improve the diversity of recommendation while considering the accuracy of recommendation.Graph computing is a common knowledge system for studying recommendation algorithms.Based on the bipartite graph,this paper will establish a three-pair bipartite graph with three types of data to study the pairwise relationship between the three types of data,and quantify each relationship,and then build a triangular model based on this.At the same time,this paper will prove the key calculation process in this model from a mathematical point of view,so as to show that the hybrid strategy proposed in this paper is reasonable.Finally,realize diversified personalized recommendation.The main work of this paper is as follows:1.About data:(1)Choose public data sets,such as MovieLens,Netflix,etc;(2)In addition to the two regular data element of users and goods,the “third dimension” data will be added in this paper;(3)Sort each user's behavior data in ascending order by timestamp.2.Constructing triangular model: First,how to construct the triangular model is introduced.Secondly,how to quantify the relationship between data in the model.Finally,from the point of view of mathematics,the calculation process and diversity are proved to ensure the rationality of the method.3.Model test indexes: The common model test indexes are introduced,and MAP index,one of the common evaluation indexes,is discussed.4.Analysis of experimental results: The personalized recommendation algorithm based on hybrid strategy proposed in this paper aims to improve the diversity of recommendation while considering the accuracy of recommendation.Since HC(HeatConduction)algorithm focuses on the diversity of recommendation,this paper will:(1)compare with HC algorithm;(2)analyse the recommendation results of the proposed algorithm from the perspective of data;(3)compare with other recommendation algorithms.The personalized recommendation algorithm based on hybrid strategy proposed in this paper is parameter-free.The algorithm has been proved to have better recommendation performance through the test of public data sets and common test indexes.The research in this paper is complementary to the research in the field of recommendation algorithm at the present stage,and is a brand new research idea,which is of great significance for the application of recommendation algorithm in practice.