Multi-objective Learning Path Optimization Method Based On Event Hypergraph

With the development of Internet and the rapid popularization of online learning,more and more learners learn knowledge through online learning platform.However,massive online learning resources not only provide effective knowledge for learners,but also cause serious knowledge overload and learning confusion.Personalized learning recommendation,as an effective method to solve the above problems,can meet the specific knowledge needs of learners.In the traditional offline teaching mode,learners have some problems in the learning process,such as unclear network structure of course knowledge points,lack of logic of knowledge system,etc.In order to help learners better learn and master fine-grained knowledge points and generate learning paths that conform to their individual characteristics,it has become a hot research issue at present.Learning path optimization aims to generate and optimize the knowledge learning order that best meets the learners’ knowledge needs based on learners’ learning behavior model and curriculum knowledge structure characteristics.The historical behavior data of learners in the process of online learning can reflect their learning preference and learning situation from the side.Learning behavior model is a general method to mine learners’ learning behavior preferences.This method can reflect learners’ learning situation from a deep level and further explore the problems existing in their learning.Therefore,this paper proposes a multi-objective learning path optimization method based on event hypergraph,which mines the dynamic learning preference behind the learning behavior and predicts the next behavior.Finally,it recommends the learning path that meets the learning preference and needs for learners by constructing a multi-objective optimization model.Firstly,this paper gives the definition of learning events in online learning and constructs a learning event hypergraph to model learning behavior.Based on the historical learning behavior data of learners,by analyzing the characteristics of their behavior data,the learning behavior data is divided into multiple learning events with time attribute information,and then an individual learning event hypergraph is constructed with learning events as hyperedges,and then the individual event hypergraph is completed by using all learners’ learning event hypergraphs to obtain a relatively complete representation of learners’ behavior characteristics.Secondly,a Learning Behavior Prediction Model based on Event Hypergraph(EHLBP)is constructed in this paper.Based on the constructed event hypergraph structure,by using hypergraph neural network and self-attention mechanism with time and position signals,the dynamic learning preferences of learners are captured and characterized,and then the next learning behavior of learners is predicted.Finally,this paper constructs the Multi-objective Learning Path Optimization Model(MLPOM),which generates an initial learning path with learning resources as nodes based on the learning behavior prediction model,extracts a subgraph composed of knowledge points contained in the learning resources in the initial learning path,selects four aspects of learning difficulty,importance,learning cost and learning experience as optimization objectives,and then constructs a multi-objective optimization model by using linear weighting method to transform the multi-objective problem into a single-objective problem,constructs a difference function among knowledge points,and solves it based on improved BPSO algorithm Finally,the learning path with fine-grained knowledge points as nodes is recommended for learners,which meets their personalized dynamic learning preferences and knowledge needs.