Research On Learner Classification And Profile Based On Key Word Extraction And Collaborative Filtering

The rapid development of the Internet and multimedia technologies has given birth to new teaching patterns.The combination of the Internet and education has generated a large amount of user data.The mining of these data can not only help students to find gaps and fill gaps,but also guide teachers to make differentiated teaching plans based on the different characteristics of students,which significantly improves teaching and learning efficiency.In this thesis,the evaluation data of data structure course from ‘Pintia' programming teaching assistant plateform is used as research sample.Aiming at the features of rich sources of problems,diversification of learners,different practice habits and training goals,this study focuses on solving the following problems.First,how to identify the similarity between the problems based on the text information embeded on the original problems.Second,how to classify the learners based on their answering behaviors and profile for the learner clusters.Third,how to refine the information of the existed problems based on the profiling of learner clusters and provide learners with suitable problems correspondingly.In order to effectively solve the above problems,the main contributions of this article are as follows:(1)An improved problem clustering method ProSet-Detc based on uncertain graph is proposed.Using this method to perform clustering induction on problems,compared with traditional text clustering methods,better performance is obtained with indicators of precision,recall,and F-value.(2)A learner clustering method FMT-RE based on hybrid collaborative filtering is proposed.When performing learner cluster detection,it effectively reduces the running time and improves the clustering quality of the learner clusters.(3)A profile construction method of learner cluster is proposed,and the method of difficulty amendment of problems and multiple-mode recommendation for different types of learners based on the learner cluster profiling are proposed.As a result,the auxiliary performance of teaching assistant platform in the training process is improved significantly,and it also provides better technical support for the online open courses.