Research On The Prediction Model Of Teachers’ Online Education Satisfaction Based On Ensemble Learning

The outbreak of covid-19 in 2020 has brought huge challenges to the entire education sector.Under the call of the Ministry of Education to "suspend classes and not stop learning",primary and secondary schools in various provinces across the country have converted offline teaching to online teaching.Compared with classroom teaching,although online teaching has the advantages of breaking through time and space constraints and sharing high-quality teaching resources,it also increases the requirements for teachers and students to use multimedia teaching equipment,and lacks in interactivity,which leads to a decline in the online education experience and satisfaction of teachers and students.Therefore,improving the online education experience of teachers and students by constructing a prediction model of teacher and student satisfaction has become a research hotspot of scholars at home and abroad.Through a full investigation of the literature,we found that the current research on online education satisfaction prediction mainly has the following problems: first,the sample size is small,the feature dimension is low,and the factors that have potential impact on satisfaction have not been fully collected;second,the model has low complexity and cannot fully approximate complex nonlinear relationships;third,feature engineering is limited,and the excitation information of different categories of features to target variables is not fully mined.Based on this,this paper makes the following research:1)We designed a questionnaire for teachers’ online education with four different question categories,including teachers’ basic attributes,teachers’ online teaching behavior,teachers’ online teaching experience,and teachers’ willingness to continue using the online teaching model,and collected nearly half a million teachers’ experience data during online teaching for follow-up research and analysis.The relevant data in the online teaching process is used for subsequent research and analysis.2)Based on Res Net-18,a teacher satisfaction prediction network model(SAFO-Net)is designed,which shows performance advantages in both horizontal and vertical comparisons,and by embedding the attention module based on sparse autoencoder in the model to further improve the model performance.3)Based on the SAFO-Net model,a satisfaction prediction framework based on ensemble learning strategy and feature classification pre-training is proposed.The framework can extract and integrate the information contained in different categories of features in the teacher data,which not only significantly enhances the prediction accuracy of satisfaction,but also solves the problem of low prediction accuracy when the model deals with uneven sample data,and the effectiveness of the framework is verified by experiments.4)After completing the research on the teacher satisfaction prediction model,a safe and efficient model and data sharing scheme is designed based on the elliptic curve encryption and proxy re-encryption mechanism for the model data transmission and sharing link in the application process of the teacher satisfaction prediction model.After proving the correctness of the scheme,the safety and efficiency of the proposed scheme are demonstrated through theoretical analysis and experiments.