Hybrid Model Recommendation Algorithm Based On Knowledge Graph Application In Graduate Management Systems

As for the application goal of "educating students with three abilities" in college education,it mainly takes the thought of socialism with Chinese characteristics in the new era advocated by the General Secretary as the guidance direction,penetrates the basic educational task of cultivating virtues and cultivating people into each stage and content of college education,combines diversified educational methods and various teaching approaches with the perfect ideological education system.To promote the attainment of ideological education and the development of students’ comprehensive ability of education goals.However,with the rapid development of Internet technology,the information data in the network presents an explosive growth,leading to the problem of information overload becoming increasingly serious,and the number of core literature also presents an explosive growth.In the case of such massive data,it is not easy for graduate students and teachers to quickly find the literature materials they need,which also has a negative impact on the daily learning efficiency and quality of graduate students,which runs counter to the concept of educating people with three abilities.In this paper,the function of the existing graduate management system is deeply studied,and combined with the status quo of the "three complete education" work in colleges and universities and the areas needing improvement,a hybrid model recommendation algorithm based on knowledge graph is proposed and applied in the graduate management system,which solves the problem that the graduate students cannot find the required literature quickly to a certain extent.This paper studies from the two aspects of constructing the knowledge map of literature information and improving the personalized recommendation algorithm.The main work of this paper includes:The first is to build the literature atlas data set.This paper integrates four data sets,and makes the relevant data of the literature more accurate and comprehensive through the improved TF-IDF algorithm,hard alignment and other methods.At the same time,historical user data is added to pave the way for the subsequent recommendation algorithm.Secondly,it innovatively proposes a recommendation model RABHM based on knowledge graph,which is mainly divided into three parts: literature representation learning module.In this module,two preference diffusion components are introduced,namely local diffusion and hierarchical propagation.Through the diffusion of neighbor nodes and the propagation of higher-order information,literature node representations containing rich knowledge can be learned.In the user history representation module,this paper fully considers the literature data of user history clicks,and introduces the time attenuation signal into the sequence model to learn the user history representation.In the user preference module,this paper combines user historical literature data and literature representations rich in knowledge,learns sufficient user representations,and makes click probability prediction with candidate literatures,so as to dig out literatures that users may pay attention to.Third,the literature recommendation module is designed and implemented,and applied in the graduate management system developed in this paper to make up for the shortcomings of online deployment of "three full education",realize the interconnection,interworking and complementation of the online and offline third class,and promote the comprehensive development of graduate studies,personality and ideology..