Research On Joint Recommendation Of Knowledge Points And Learning Partners Based On Tensor Decomposition

The modernisation of higher education is inseparable from the high-quality development of online learning platforms,which are still plagued with problems such as cumbersome learning resources,a single way of acquiring knowledge and low student activity.To help students improve learning effectiveness and learning engagement,this paper proposes a knowledge recommendation algorithm and a joint knowledge and learning partner recommendation algorithm based on student learning behaviour data,knowledge quiz data and interaction behaviour data from online learning platforms,with the following main work:1.This paper proposes a knowledge recommendation algorithm based on tensor decomposition and Transformer reordering to address the limitation that existing knowledge recommendation algorithms do not effectively consider the learning order of knowledge points and the integrity of heterogeneous data.First,the student tensor,knowledge point tensor and interaction tensor constructed from the heterogeneous data of the online learning platform are fused and simplified into an aggregated tensor;Secondly,the tensor-based higher-order singular value decomposition is used to analyse the aggregated tensor in a multi-dimensional way to obtain the student personalised feature matrix and the initial recommendation sequence of knowledge points;Then,the potential embedding matrix of knowledge points is obtained using Transformer by combining the initial recommendation sequence of knowledge points and the information on the learning order of knowledge points;Finally,by fusing the potential embedding matrix of knowledge points and the matrix of students’ personalized characteristics,a Top-N list of knowledge points is generated to meet students’ personalized needs.The experimental results show that the proposed algorithm outperforms the classical algorithm in terms of evaluation metrics such as normalised discounted cumulative gain and mean inverse ranking.2.A joint recommendation algorithm for knowledge points and learning partners based on tensor decomposition is proposed to address the limitation that existing learning partner recommendation algorithms do not take into account the influence of other recommended contents and student interaction information.First,a new fifth-order aggregation tensor is obtained by tensor simplification of the constructed composite tensor,and the tensor expansion and singular value denoising of the aggregation tensor are performed to extract personalized features of students;Secondly,the student personalized characteristics matrix is tensorized and the student personalized similarity matrix is obtained based on Manhattan distance;Then,based on the online learning platform to obtain student-student interaction data to calculate student common interest similarity,and use it as a correction factor to improve the student personalized similarity matrix to generate the student personalized interest similarity matrix;Finally,based on the student knowledge point recommendation sequences obtained from the first research content,the similarity of the knowledge point recommendation sequences between two students is calculated using the improved Dynamic Time Warping algorithm,and after weighted fusion with the personalized interest similarity matrix,spectral clustering is used to generate a list of study partner recommendations for the target student that match the student’s friendship intention.The experimental results show that the proposed algorithm has a good performance in terms of Precision and Recall.The algorithm proposed in this paper can be used as a teaching aid module in an online learning platform.By comparing the recommended sequence of knowledge points for different students,teachers can gain insight into students’ interests and learning dynamics,and then optimise the course design.Students can find their own learning pace and motivation based on the recommended knowledge points and learning partners to improve their learning efficiency and engagement.