Research And Implementation On Behavioral Engagement Evaluation Model Of Online Learning Based On Machine Learning

The development of online education has changed the limitations of traditional education,enabling learners to flexibly control their own time whenever and wherever,and teachers can also to teach at home.However,online education has some disadvantages,such as high flexibility makes some learners less involved,which leads to poor learning effect.At the same time,due to the lack of supervision and the lag of assessment,some learners may start to "study hard" when the class is about to end,resulting in the learning effect cannot be guaranteed.With the development and application of educational data mining and learning analysis technology in teaching,the learner data stored in the platform has increasingly become the focus of scholars' research.By analyzing all kinds of learners' data in the platform,it can provide assistant for learners' self-adjustment,teachers' timely intervention and personalized learning.Starting from the learner's behavioral engagement,this study explores the method of timely evaluation of learner's behavioral engagement in online learning environment and builds a system.Firstly,by reviewing related literature,this study summarizes study into the teaching of research,combing behavioral engagement in traditional teaching and the evaluation method and effect of online teaching,it is concluded that the current online learning behavior in evaluating existing limitations.In order to timely review of learners in online learning behavior,I set up behavioral engagement evaluation system.At the start of the course in week 8,the evaluation results can timely feedback to the learners and teachers,help learners to adjust in time,teachers teach better.Secondly,in order to realize timely evaluation and measurement of online learning behavioral engagement,this study sorted out the evaluation indicators of online learning behavioral engagement and built a five-dimensional indicator model,including:Participation,focus,persistence,interaction and performance efforts,there are 13 indicators,and using the data mining technology and SPSS analysis of 13 indicators,analysis after rejecting a null value index and high correlation index,eventually determine the proportions of the video watched and chapter test finished,high traffic keep,the number of ruminant is more than 100% and the average video viewing time as the main indicator of this research,and according to the five indicators for clustering analysis,clustering quality inspection,and a single indicator of grade correlation analysis,verify the validity of the five indicators.Finally,using Python machine learning algorithm to build logistic regression,random forest,support vector machine SVM and integration method Stacking models,after training found that the first 8-13 weeks of the effect is poor,the reasons are as follows: data set imbalance,indicator selection is not appropriate or individual classifier selection is not appropriate;After the data set is balanced by the over-sampling method,it is found that the effect of the model is improved in 8-13 weeks after re-training.By comparison,it is found that the effect of the random forest model is better than that of the other three types of models.Therefore,the random forest is adopted as the main algorithm for the development of behavioral engagement evaluation system.Develop the system using Flask Web and store the data in a local SQLite database.The system is divided into student side and teacher side.From the eighth week,learners can check the specific data of their weekly behavioral engagement,their class ranking and the evaluation results of behavioral engagement.Teachers can comprehensively view the behavioral engagement data of all learners in the class as well as the correlation graph with their performance.