The Design And Implementation Of Early-warning System Based On Machine Learning

With the development of intelligence and information technology in online education industry,learning early warning technology has become a hot field of online education research and application in recent years.In this paper,two hot issues in the field of learning early warning,score early warning and drop out early warning,are studied.There are some problems in score early warning,such as large particle size and low precision.In this paper,a learning early warning system based on machine learning is designed.According to the users relevant behavior data,we can establish score early warning model and withdrawal early warning model through data mining method.According to the early warning effect of the model,we can send relevant early warning results to students and teachers,which helps to improve the quality of teaching.The main research work of this paper is as follows:(1)This paper summarizes the development process and current situation of learning early warning system at home and abroad,introduces the current development and application of machine learning technology,studies the related papers of learning early warning model based on machine learning,and puts forward the design requirements of learning early warning system for score early warning and withdrawal early warning scenarios.(2)Aiming at the task of learning score early warning in learning early warning,based on the data in the mixed learning environment of online learning platform and offline classroom of software engineering course in a university,this paper constructs fine-grained features based on knowledge points and question types,Based on the neural network model,"knowledge point feature network + other feature network" and "question type feature network+ other feature network" are designed to deal with two kinds of features respectively,which solves the problem of large early warning granularity in the score early warning problem to a certain extent.(3)Aiming at the task of early class withdrawal warning in learning early warning,based on the students course selection and learning data of a university teaching network,this paper abstracts the course selection scene as a class withdrawal classification problem,and designs a model fusion strategy based on xgboost and lightgbm by using the good effect of integrated learning on numerical type classification problem,To a certain extent,it solves the problem of low precision in the early warning of class withdrawal.(4)A learning early warning system is designed for the situation of score early warning and withdrawal early warning,and has been applied to the actual course.By analyzing the functional and non functional requirements of the learning early warning system,the overall design and detailed design process of the system are described through detailed charts.On this basis,the score early warning and withdrawal early warning modules are completed.After learning the early warning system,the system is tested,and the test case results show that the functions of the system are normal.In the score early warning task,the accuracy of the model is 85.81%,the mean square error is 89.78,and the Pearson correlation coefficient is 0.89,which is better than the same type of model.In the task of early warning,the accuracy rate is82.57%,the accuracy rate is 74.23%,and the recall rate is 62.31%,which is also better than the same type of model.It is verified that the score early warning model and the early warning model have good performance in their respective tasks.Through the above research work,the application of machine learning in learning early warning is realized in this system.