Research On Clustering Of Learners Based On Data Mining ——Take The Mongolian MOOC Platform As An Example

The rise of online education allows learners to use the Internet to learn courses anytime and anywhere.As a result,a large amount of behavioral data has been generated on the MOOC platform.These data seem to be messy and disorderly,but in fact they hide learners’ learning behavior characteristics.If learners are clustered and analyzed according to their behaviors,not only can they understand the differences in learning behaviors between different learners,but also help teachers adjust the teaching process in time.In addition,the text of the discussion area reflects the dynamics of the course,and is an important manifestation of teacher-student interaction and student-student exchanges and collaboration.Through analysis,topics of interest to different learners can be obtained.However,the current discussion forum in most platform is only for discussion and discussion does not play a practical role.Therefore,research focuses on learners’ online learning behaviors,clusters learners according to behavior indicators,compares differences in online learning behaviors,hot topics,and learning emotions among different learning groups,summarizes learning characteristics,and provides reference suggestions for improving online education.Research uses the online behavior data and discussion forum texts of the three courses of "Middle School Information Technology Teaching Skills Training","Mongolian Folklore" and "Introduction to Computer" in the Mongolian MOOC platform as the research object.And using cluster analysis,text analysis,and comparative research method to carry out the research work.The main work of the research is divided into four parts: First,establish behavior analysis indicators and define learning groups.On the basis of analyzing the actual effective learning behavior of learners in the online learning process,a behavior analysis framework is constructed.Using hierarchical clustering method and K-Means clustering algorithm,learners are clustered into four categories: "normal learners","passive learners","interactive learners" and "active learners" based on behavior analysis indicators.Second,analyze the differences in online learning behaviors among different learning groups.First of all,conduct a statistical analysis of learners’ learning behaviors to understand the overall situation of learning behaviors of different learning groups.Then use the analysis of variance method to compare and understand the specific differences in time investment,learning effectiveness,learning interaction and learning enthusiasm of different learning groups.Third,analyze the similarities and differences of hot topics among different learning groups.On the first,the research pre-processes the discussion text,then uses the TF-IDF algorithm to extract keywords,clusters them according to the semantic similarity of the keywords,finally get the hot topics of each type of learning group,compare and analyze the similarities and differences of the knowledge content.Fourth,analyze learning attitudes of learning groups.Above all,the method of sentiment analysis is used to explore the distribution of learning emotions of different learning groups,next is to use questionnaire survey and interview methods to analyze the satisfaction of learners and teachers,and explore the specific factors that affect online learning.In the end,the integrated research results provide useful suggestions on how to improve online teaching.The main conclusions of the research are as follows: First,there are a large number of learners participating in online learning and their styles are very different.By analyzing the behavior differences between different learning groups,teachers can understand the learning situation of all learners in a short time and take timely measures.Thereby improving the learning effect.Second,the more even the distribution of learners’ online learning behaviors,the more favorable it is to achieve good learning results.Learners who only invest in tests or interactions fail to achieve the desired learning results.Third,the analysis of teaching effects found that teachers and students all have a positive attitude towards the mixed teaching model,indicating that this model has a certain feasibility for promotion in colleges and universities.Fourth,the study found that the online behavior of learners is rich and diverse,but the current assessment and evaluation are mostly based on performance.Therefore,the evaluation mechanism in the learning process should be improved to highlight the dominant position of learners.