Research On The Prediction Of The Application Ability Of Teachers’ Digital Educational Resources Based On Machine Learning

The advantages of open sharing of digital education resources enable it to quickly respond to the needs of learners at different levels,and its effective application is also an important means to solve educational equity.However,due to the limitations of some teachers’ own abilities,the expected results have not been achieved in the use of existing resources to carry out teaching.Therefore,exploring how to realize the effective monitoring of the application ability of digital resources has become an important part of the teacher’s ability improvement project and solving the difference in education application.However,the traditional questionnaire assessment is difficult to cover every individual teacher,and the commonly used linear regression model is often the research and evaluation of a group,which cannot satisfy and support the more fine-grained individual teacher research.However,with the rapid development of computer software and hardware,computer computing power is no longer a limitation of the experiment.This paper uses machine learning related algorithms to solve the shortcomings of the linear regression model’s weak ability to fit samples,and realize the research subject from the group to personal transformation,in order to achieve the use of intelligent assessment to optimize the training program to promote the growth of teachers.Based on the above background,this article aims at accurately predicting the application ability of teachers’ digital education resources,using machine learning algorithms to realize the factor analysis of teachers’ application ability of digital education resources,and perform feature engineering based on the analysis results,and then build a model to perform application status Pre-judgment,the main work of this paper is as follows:(1)Analysis of the influencing factors of teachers’ application ability of digital education resources.After the data is preprocessed,assumptions are made on the background variables and training variables at the two levels of the teacher and the school,as well as the interaction relationship between them,and the multi-layer linear regression is used to verify the results.The application ability of digital education resources has a positive impact,and there is an interactive effect between the two levels of indicators,which also provides screening for further predictive indicators.(2)Construct a prediction model based on multi-layer linear regression,random forest,XGBoost,and LightGBM.Through experiments,the average absolute percentage error indicators of each model are 15.22%,3.96%,4.02%,and 5.13%,respectively,which verifies that the regression algorithm in machine learning is better than the traditional linear regression prediction algorithm in the prediction task in the education field.Because the structure of random forest is quite different from XGBoost and LightGBM,this paper combines these three models again and conducts experiments,and compares and finds that the multi-layer combination model based on random forest is the best choice for prediction.The average absolute percentage error The model score after the fusion of the indicators is the best 3.68%,which is better than the 3.96%of the random forest..(3)Application of prediction results of rural teachers’ digital education resource application ability.At the regional level,show how to conduct macro-monitoring and precise identification of training needs based on predicted values,so as to support the management department in making decisions on the number of training times,the selection of trainees,and the organization of training scales.Provide score rankings and indicators of various training dimensions at the individual level of teachers to assist teachers in their self-awareness,and at the same time facilitate the school to carry out differentiated management.