| Massive open online courses(MOOC)have provided users with a diverse learning experience since 2012.However,the large number of MOOC learners limits teacherstudent interaction with the lack of effective instructor intervention on students.In the context,the paper use the rich data resources in MOOC forums to explore the factors that affect instructor intervention and the factors of topics responded by teachers.And threads which require instructor intervention is predicted.Also,a new path for researches on instructor intervention in forums is provided by the big data analysis and machine learning techniques.In the study,22 topic discussion features are put forward and the prediction model of discussion-oriented instructor intervention in MOOC forums based on related literature and the learners' discussion process in MOOC forums.The specific steps are: data collection and processing,topic discussion features analysis,prediction algorithm for threads that require instructor intervention,prediction and intervention.Among them,the part of prediction algorithm for threads that require instructor intervention includes: feature selection,algorithm selection,data set split and model evaluation.Based on this,the study takes the topic discussion data(8932 topic posts from September to November 2018)of advanced mathematics(I)course on the “China University MOOC” platform as the object and perform experimental research on the model above.First,the data is obtained through a web crawler program and is extracted and processed according to the machine learning feature engineering method,then algorithms such as logistic regression,decision tree,random forest,and support vector machine are selected to establish the prediction of teacher intervention topics.Then,the "best" intervention strategies for threads which require teachers' response is explored.Specifically,it includes: teacher-student behavioral transition patterns;correlation between topic types,teacher intervention types,intervention time,and learners' discussion levels in follow-up topics;characteristic of topics with teacher intervention in MOOC forums;advice for discussion-oriented teacher intervention in MOOC forums.Results show that:(1)The topics worthy of teacher intervention in the MOOC forums have characteristics of high attention,low response,and basic knowledge;(2)Among the four types of algorithms,the prediction model based on decision tree algorithm works well,with the accuracy close to 0.8.The optimal feature set conducted is forum type,quality of question description,richness of question,learners' participation in discussion,test-related question,time of question and importance of knowledge points;(3)The instructor intervention brings the learner's interaction depth to the discussion topic.Different types of topics often correspond to different optimal intervention times and intervention types.According to the results,some suggestions are put forward.(1)Set priorities or establish a reasonable ordering method for topics that require teacher intervention in the MOOC discussion forum.For similar topics in the MOOC discussion forum,set up a search function or text similarity calculation,and recommend similar issues and answers obtained in the past.When providing topical intervention recommendations for teachers,the topic of secondary intervention should be taken into account.In addition,the division of labor in multiple teaching roles should be considered to recommend topics for users with different teaching roles.(2)Instructors should provide multiple interventions on the topic of discussion,and multiple types of intervention should be conducted in parallel to guide learners to think deeply,inductively transfer,and create a space for learners to collide with each other.At the same time,pay attention to the emotional interaction between teachers and students,and give learners more Positive feedback and expectations,design innovative open-ended discussion topics,and improve student participation. |
PDF Full Text Request