Improvement Of Recommendation And Prediction Algorithm Integrating External Information

In the era of big data,data is constantly emerging.How to efficiently mine the value of data and how to make full use of the information contained in data to serve individuals and promote the development of society are important issues in the era of big data.Recommendation algorithms and link prediction are products in the context of the era of big data,which are used to mine the value information of data.Today,massive amounts of data come from different systems,each system is interrelated,and the information in each system affects each other.Therefore,how to put the data in different systems under a unified system framework is an important topic at present..In the same way,the recommendation system and link prediction should also consider the influence of other systems,and use the auxiliary information of other systems to improve their performance.With the development of recommendation system technology,the recommendation algorithm is also constantly developing,and the recommendation algorithm based on the user-product bipartite structure has become a research hotspot in the field of recommendation algorithm.At present,most of the improvements of recommendation algorithms based on material diffusion are improvements on a single user-commodity binary network system,and no information from other systems is introduced to assist the recommendation.In the era of big data,various systems are interrelated.When studying the improvement of the material diffusion algorithm,it should not be limited to using a single user-product interaction information to recommend products.Based on this,this paper proposes a substance diffusion recommendation algorithm coupled with social network information to effectively couple the social information in the social system to assist the substance diffusion recommendation algorithm for recommendation,so that the recommendation effect is better.In complex networks,link prediction is another important research topic.The algorithm for predicting by defining the similarity index of nodes is an important method in link prediction.At present,most of the node similarity indicators used for link prediction only consider the link structure information of the link prediction system itself,ignoring the attribute information of the chain edge in the network,and the attribute information of the chain edge is not negligible.The similarity index of the nodes learned with the attribute information of links can better reflect the attributes of the nodes.This paper proposes its own research method based on this topic:a new link prediction similarity index based on the graph embedding algorithm,which can couple the sign attribute information of the link.The newly proposed algorithm couples more attributes to the node similarity index,which makes the new node similarity index perform better in link prediction experiments.