Research On Learner Topic Mining And Learner Retention In The MOOC Environment

MOOCs have recently experienced rapid expansion.It is a new innovation through deep integration with emerging information technology and education.It offer extensive diversity of learning to improve the ability,expand learners international vision and further promote all people for lifelong learning.Despite the momentum for MOOCs,MOOCs learners have a high dropout rate and a large number of learners can?t complete the course.This phenomenon has been questioned and criticized by the industry,which has also gained significant attention among educators.The indicators of low completion rate and low retention rate is more relevant to traditional education evaluation,it is not the most appropriate measure to evaluate the effectiveness of MOOCs,but the low completion rate does raise questions regarding learning quality and pedagogical practices.Because MOOC learners? main goal is not to complete the entire course,tend to regarded MOOCs as be a learning experience,or to selectively study part of the whole course when needed,so the completion rate and retention rate indicators in the pedagogy perspective is not the best way to measure the effectiveness of MOOCs.In the view of information system,regardless of whether learners complete an entire course,MOOCs learning is regarded as the adoption and continuous usage of MOOCs,and it is more suitable to reveal the effectiveness of MOOCs and its sustainability.Based on the perspective of pedagogy and information system,learners retention includes two stages: initial adoption and continuous usage of MOOCs.Our main aim is to understand how MOOCs?teaching support service can meet the needs of learners,bring the wonderful initial registration experience of learners and finally promote the continuous learning behavior of the learner.Two-stage learners retention reseach to make up for the insufficiency of the “only registered” experience in the existing literature.Considering the social characteristics of MOOCs,the specific learning behavior of MOOCs learners is extended to: learner registration,continuous learning and dissemination of MOOCs resource with social media.These learning behaviors form a positive learning cycle“Social Media Dissemination-> Learner Initial Registration-> Learner Continuous Learning”,the social media dissemination mechanism leads more learners to participate in MOOCs learning,and learners after using MOOCs tend to share their experiences and gains with social media,thus promoting the healthy and MOOCs?sustainability.We introduces learning theory,cognitive load theory and cognitive theory of multimedia learning,which involves many disciplines such as educational psychology,management science,computer science and cognitive neuroscience,integrated text mining,machine learning technology,statistical modeling and so on.With the problem of MOOCs learners retention,the MOOCs learning process is regarded as a dynamic learning cycle “Social Media Dissemination-> Learner Initial Registration-> Learner Continuous Learning”,and the specific work of this paper includes the following aspects:Firstly,the research on MOOCs topic mining for social media reviews provides a reference factors affecting learner retention.MOOCs are characterized by social media,and learners prefer to share and disseminate MOOCs resources through social media,and social media user reviews provide important and unique sources for MOOCs research.We collected MOOCs-related data from social media,including the user-generated content of six MOOCs platforms such as Coursera,ed X,Udacity,Xuetang X,icourse163 and MOOCs College,and analyzed the topic heat of MOOCs and the public trend.In order to deeply understand that learners?needs,we use the text mining technology to identify MOOCs topic.We propose the GPLSA method,which combine the probabilistic topic model and word co-occurrence relations.The specific implementation step comprises the PLSA algorithm to preliminarily identify the subtopics,remove the common background word,merge similar redundant subtopics and update topic features.The algorithm evaluation show that the results of the GPLSA proposed by the present study is higher than the traditional K-means and LDA,and the generated topic quality is higher.The GPLSA method effectively identifies learners?needs subtopics,including quality MOOCs platform,practical course resources,teaching teachers,certificate motivation,emotional interaction and various confusion in learning process,which provides a reference factors affecting learner retention research.It is worth noting that some learners will post for help,such as choice confusion caused by rich courses,delay of video playing,setting of chinese subtitle,use of mobile phone client,frustration of difficulty in course and so on,which reflect that learners are not able to obtain the teaching support provided by the MOOCs platform in time,then seek help from the social platform,so MOOCs pedagogical support services also need to be improved.Secondly,the situational cognition of MOOCs learner initial registration.Due to the lack of research on learners who quit halfway,even registered only,we uses two-stage learners retention reseach,namely initial adoption and continuous usage of MOOCs.Initial adoption is considered as the decision process of the learner initial registration,which is very important and has a great influence on the learners?continuous learning.Based on the cognitive load theory and the cognitive theory of multimedia learning,from the perspective of cognitive neuroscience we puts forward a situational cognition model with the characteristics of the primary emotional information and the secondary emotional information.We specifically discusses the information presentation of MOOCs service products,including no teacher presentation mode and teacher presentation mode,which influence the learner's attention,emotion and learning effect,combining the online course browsing experience evaluation with the objective eyetracking technology,observed the cognitive track of the learner and disclosed the situational cognition mechanism which promoted or impeded the learning experience.The results show that:(1)There are “hot potato” behaviors in both presentation mode,the learner?s attention has avoided the picture region;(2)The teacher image presentation significantly affects learners? perceived usefulness and visual attention,however,that ability of the participant's information recall is not better or worse than that of no teacher presentation;(3)The teacher?s presentation mode changes the learner?s browsing behavior and guides the learner to participate more actively in the cognitive activity.According to the above conclusions,we can further develop a situational cognition mechanism to promote the learning experience.In the MOOCs learning scenario,teachers? presentation meets peoples?emotional needs,conforms to the design principles of multimedia learning cognitive theory,and increases learner germane load,which effectively guide learners to invest more cognitive resources into cognitive activities related to learning based on selective attention regulation,thereby stimulate the learners? persistent secondary emotional reaction,improve the cognitive experience of the learners,and increase the likelihood that learners continue to learn.Finally,dynamic process modeling of MOOCs learner's continuous learning behavior.A good experience for learners to adopt MOOCs will increase the possibility of visiting the MOOCs site again,therefore,it is necessary to consider how to retain the existing learners and form a positive learning behavior cycle in the second stage of learners retention.MOOCs have the characteristics of learning process.According to the learning framework of Knud Illis,learning includes two processes,one is the learning acquisition process,most of the MOOCs learners are driven by internal motivation,so as to acquire knowledge and improve their abilities,we defines this process as the process of learner's ability change;the other is the interaction process between the individual and the external environment,wherein the external environment factor is specifically defined as website,course,peer effect and teacher-student interaction,and the interaction process is essentially the dynamic process of a series of rational learning decisions,which is the result of interaction between individual ability and external environment factor.Under the environment of MOOCs,the difference of learners? ability is increased,and there may be completely different cognitive processes for the same pedagogical environment,therefore,we introduced expertise reversal effect to analyze the dynamic change rule of learners? ability and the influence of interaction process between learners? ability and external environmental factors on learning decision.Considering the non-observable characteristics of the ability state and the learning characteristics of the two processes,referring to the hidden Markov model,the learner ability state is taken as the implicit parameter,and the interactive result of the individual and the environment,that is,the learning decision is regarded as the observable parameter,we model the dynamic process of learners? continuous learning.The model parameters are estimated by maximum likelihood,and the results show that:(1)For different ability the influence of external factors(website,course,peer effect and teacher-student interaction)on learning decision is different.Peer effect has positive influence on learning decision of learners with different abilities,learning decision of high learners is influenced by factors such as websites and courses,low learners are more willing to participate in teacher-student interaction activities.(2)Learning actively promote the improvement of learners?ability.However,high and low learners have differences,high learners have accumulated advantages,more likely to succeed and progress.Further analyse the causes of accumulation advantages,because high learners show more active learning behaviors,therefore,it is more likely to maintain their own accumulation advantage and status,and finally identify the evolution path and rule of the learners?ability state.According to the above conclusions,in the learning environment of MOOCs,try to avoid monotonous one-size-fits-all pedagogical method,and provide effective intervention method according to learners? ability and demand.Compared with the existing research results,the innovations of this paper are as follows:First,based on MOOCs social media data mining the learner's demand,we broken through samples size limitation of the traditional survey and interview.We propose the GPLSA method,which combine the probabilistic topic model and the graph mining method,integrating semantic information and word co-occurrence relations and avoiding the loss of semantic relationship between words in original documents,thus effectively improve the quality of topic detection.The GPLSA method not only makes a new exploration on the MOOCs topic detection method,but also expands the research on MOOCs learners? demand mining in China.Second,in the learner initial registration phase,under the guidance of cognitive load theory and the cognitive theory of multimedia learning,we proposes learners situational cognitive model based on cognitive neuroscience,which provides a new perspective for the cognitive process of MOOCs learners? initial experience,and reveal the situational cognitive mechanism that promotes or hinders learning experiences.The eyetracking technology not only avoids the subjective influence of the participants,but also enriches MOOCs research methods.Besides,the study of the initial registration process of the learner makes a beneficial supplement to the cognition of the “only registered”.Third,in the learner continuous learning phase,referring the hidden Markov model,the learner's ability state is taken as the implicit parameter,and the interaction result of the learner's ability and the external environment factor,that is,the learning decision is regarded as the observable parameter,we model the dynamic process of learners? continuous learning.We extend the HMM learning decision model,the computing problem of the ability state transition probability is reduced to the construction of an ordered Logit model,and the learning decision probability calculation is transformed into a binary Logit model.The integration of learning theory frame and hidden Markov model,not only enriches MOOCs research,providing new ideas and methods for analyzing the persistence problem of MOOCs,but also expands the application of HMM model in the management field.