Achievement Prediction And Target Pushing In Personalized Learning Systems

As the demand for online learning grows among learners,more and more educational resource platforms and personalized learning systems are being developed.In the face of the explosive growth of learning resources,it is difficult for learners to find learning resources that match their cognitive level and current knowledge system,and it is difficult for teachers to tailor their teaching to their needs in a personalized learning system.To address the problems of high course dropout rates and unsatisfactory course grades in online learning platforms,we propose a grade prediction model to track students’ learning by analyzing the factors influencing academic performance;by analyzing the difficulties of course learning effects and learning goal recommendations in current online learning systems,we propose a learning goal push method based on personalized course knowledge points to compensate for the lack of face-to-face communication and guidance from teachers in online learning systems and improve learning efficiency,as follows:(1)A variety of classification algorithms were used to explore the mechanisms that influence the relationship between students’ characteristic factors and their course performance,starting from 21 characteristic factors in four areas: students’ personal characteristics,sociodemographic characteristics,learning engagement,and learning environment.Analyzing important features that influence trends in student course test scores sets the stage for a performance prediction model that can capture dynamic changes in student test scores.(2)Student data were organized and pre-processed to propose a multi-hidden layer improved neural network grade prediction model,which considers the association between feature factors and grade impact,and selects five important feature factors as model inputs.The proposed model is able to focus more on the association between multiple features and scores,improving the generalization as well as the learning ability of the model.The experimental results show that the proposed model in this paper has 9% improvement in accuracy,precision and F1 value,the lowest average absolute error performance index and the least volatility compared with other machine learning algorithm models.(3)We propose the construction method of personalized cognitive structure,which represents the cognitive structure of learners by means of attribute expansion;then we divide the personalized cognitive structure into hierarchical levels according to the characteristics of sequential and easy to difficult knowledge points;then we comprehensively analyze the hierarchical information and association strength of nodes in the cognitive structure,and propose the calculation method of learning influence degree between nodes;finally,we combine the situation of knowledge points already mastered by learners and the influence degree of knowledge points to calculate the learnable expectation value of new knowledge points,and gradually push the appropriate new knowledge points for learners actively.