Research On Prediction Model Of Class Withdrawal In Online Teaching Platform

In recent years,under the background of global education modernization,large-scale online open curriculum(MOOC)as a new online teaching model has attracted extensive attention.Compared with the traditional offline teaching mode,MOOC platform not only provides low-cost and high-quality curriculum resources,but also advocates students’ autonomous learning according to their own learning needs.However,due to the lack of supervision in the learning process and the lack of timely communication between teachers and students,MOOC platform has a high withdrawal rate,which not only affects the normal development of teaching activities,but also becomes the biggest obstacle restricting the sustainable development of MOOC platform.In the online teaching environment,a huge student group and a variety of MOOC courses together constitute a very complex learning system.How to analyze students’ withdrawal factors from complex teaching data resources is the primary goal of many researchers.Taking the online MOOC platform of the school as the research object,this paper makes an in-depth analysis of MOOC withdrawal from multiple angles around the students’ learning behavior data,and completes the application of deep learning method in MOOC withdrawal prediction from the practical level,as well as the training,integration and improvement of basic classification model.By summarizing the results of experimental comparison,this paper deeply discusses the MOOC withdrawal prediction model and its application strategy proposed in this paper.The main research work of this paper is summarized as follows:(1)Aiming at the problem of data sparsity in MOOC teaching data,the feature engineering method is applied to educational data mining to analyze the learning behavior of MOOC students from a new perspective.According to the overall and phased learning state of students,various behavior characteristics are deeply explored,and the two behavior characteristic data of traditional characteristics and time series characteristics are extracted to provide data support for the modeling of withdrawal prediction model.(2)According to the time characteristics of MOOC course teaching and students’ course learning,a CNN bilstm withdrawal prediction model based on deep learning is proposed from the perspective of time series prediction,which makes a fine-grained analysis of students’ phased learning state.At the practical level,the effectiveness of CNN bilstm model in MOOC withdrawal prediction is verified from three aspects: single course model prediction,model reuse and model effectiveness.Compared with other benchmark models,the prediction accuracy of CNN bilstm model is more than 84%,which has a better prediction effect on the prediction of MOOC withdrawal.(3)Based on the above CNN bilstm model,a MOOC withdrawal prediction model based on multi model fusion is proposed to deeply study the MOOC withdrawal prediction from the perspective of the combination of students’ overall and phased learning state.The experimental results show that through the effective innovative fusion of the base classification model,the accuracy and AUC value of the overall model have reached more than 91%,which verifies the effectiveness of the fusion model proposed in this paper.In summary,this paper provides effective ideas and methods for the analysis of withdrawal behavior of online teaching platform from the perspective of data-driven.