Research And Application Of Student Academic Early Warning Based On Convolutional Neural Network And SVM

With the development of higher education in China,the phenomenon of skipping classes and failing to pass has begun to appear at the same time as the enrollment expansion of colleges and universities and the surge in the number of college students,which has seriously affected students' studies and the school spirit.Therefore,effective academic warning is becoming more and more important.In order to guarantee the teaching quality and urge students to study,this paper presents a scheme of academic warning which combines attendance early warning and fail early warning.The attendance of students is carried out by face recognition,and attendance warning is generated according to attendance conditions;based on the attendance status of students and historical examination results,the course of the current semester is predicted by machine learning,and fail early warning is generated according to the prediction;attendance early warning and fail early warning are used to carry out academic warning for abnormal students.The main contents of this paper are as follows:(1)Attendance early warning module based on face recognition focuses on face recognition model based on convolutional neural network.In this paper,the widely used mtcnn network is used for face detection and face alignment;the face matching algorithm,combined with the history of the development of convolutional neural networks,designed the convolutional neural network MYNet based on the 3×3 size convolution kernel.Then,the experimental data set is constructed by collecting face photos.The comparative experiments verify that the MYNet network designed in this paper has higher recognition accuracy than the classical convolution neural network with the same number of layers and its improved network.(2)Fail early warning module based on the prediction of failing course focuses on the prediction model of course failing course based on SVM.In order to avoid the impact of the default SVM super parameters on the performance of the model when using SVM algorithm to predict the students' current courses,this paper uses grid search algorithm and 3-fold cross validation to optimize the super parameters of SVM according to the actual use scenarios of the algorithm.After that,the internal relationship between courses is analyzed,and the experimental data set is constructed.Through the comparative experiments,it is verified that the SVM model optimized by grid search algorithm has higher performance.Based on the research and improvement of the algorithm,this paper designs and implements the early warning system of students' academic performance.The system realizes the students' face recognition attendance and fail forecast;realizes the attendance early warning by analyzing the attendance data,and realizes the fail early warning by analyzing the forecast data;finally realizes the academic early warning of abnormal students by integrating the attendance early warning and fail early warning,and effectively improves the students' management efficiency of teachers and counselors,as well as the students' learning quality and academic level.