Research On Learning Analysis And Personlized Recommendation Based On Online Learning Platform

In recent years,with the vigorous development of big data and artificial intelligence,the education form has undergone profound changes.The rapid development of online education has led to the emergence of various online learning platforms.While online learning platforms provide a huge amount of learning resources,how to recommend suitable learning content for students has become a difficult problem.In the field of education,the existing personalized recommendation research has many problems:In terms of research content,the current personalized recommendations in the education field aim to increase click-through rates,such as recommending courses students interested in.Such research cannot help students to strengthen their grasp of weak knowledge points and improve their academic performance.In terms of research methods,on the one hand,the traditional collaborative filtering methods cannot effectively solve the problem of data sparsity;on the other hand,content-based recommendation methods usually use simple behavioral characteristics and lack the support of professional learning analysis and pedagogy theory.Based on the real data of online learning platform in colleges,this thesis analyzes the knowledge concept map and learning characteristics,constructs a recommendation model based on neural network,and finally recommends suitable online learning resources for students to enhance their grasp of weak knowledge points.In order to explore the connections between the knowledge points behind the learning resources,this thesis constructs the knowledge concept map,which characterizes the knowledge structure of learning resources,and then uses knowledge embedding technology to vectorize entities and relationship,providing auxiliary information of knowledge structure for recommendation model.In order to take advantage of students' hidden learning behavior characteristics,this thesis analyzes students' learning behavior data and constructs high-dimensional learning behavior feature vectors based on Bloom's Taxonomy and Solomon Learning Style theory.In terms of recommendation algorithm,this thesis proposes a hybrid recommendation model CKNN(Collaborative Knowledge Nerual Network)based on multi-layer perceptron neural network,which solves the problem of data sparsity through Autoencoder,and integrates knowledge concept map and learning style characteristics to achieve accurate personalized recommendation.Finally,the Attention mechanism is introduced to optimize CKNN.On the one hand,the ECA(Efficient Channel Attention)module is embedded to scale low-dimensional and high-dimensional features.On the other hand,in order to introduce the related knowledge information,this thesis embeds the associated entities of the knowledge vector,which further improves the model performance.