Research On Personalized Learning Path Recommendation In Learning Devices

At present,there are still some problems in most learning platforms,such as unclear characteristics of learner model,inability to represent learner diversity,and inaccurate personalized recommendation results.To solve the above problems,this paper constructs the static feature system,dynamic feature system,static learner model,dynamic learner model and personalized learning path recommendation model.In the feature system,the static feature system is mainly divided into four categories,and the dynamic feature system includes three additional categories on the basis of static features,and finally expresses the features in mathematical form.Then,the static feature data is classified by K-means clustering method,and the static learner model is constructed;Combining static feature data and dynamic feature data,BP neural network is used to build dynamic learner model.At the same time,this paper uses the idea of knowledge mapping and reordering to build a personalized recommendation model.Firstly,the knowledge map of the course is constructed,and then the cognitive model is used to analyze the learners’ test results.According to the test results,the knowledge map of the course is traversed,and the knowledge map of the blind area is constructed.then.The learning ranking method is used to realize the initial ranking of the knowledge points in the blind area,and then the feature vectors of the initial sequence are input into the transformer algorithm to generate the optimized sequence combined with the learner’s personalized features.Finally,the topological sorting method is used to traverse the rearrangement sequence and the map of knowledge points in the blind area to generate the final recommended knowledge point learning sequence.In this paper,offline analysis method is used to verify the model.The results show that k-means algorithm is better than Mini batch k-means algorithm;The fitting curve,ROC curve and AUC value are used to verify the construction effect of dynamic learner model.The fitting degree is 0.88431 and AUC value is 0.901;In the verification of personalized recommendation model,precision,recall,average accuracy and average accuracy are used to compare the recommendation effect of the three recommendation algorithms.The experiment shows that the recommendation method in this paper is higher than the two comparison algorithms in the evaluation index.