Application Research Of Machine Learning In Student Pass Rate Prediction

So far,education is still one of the most important targets in the country.In the teaching process,the quality of students is mainly reflected in student achievement or student graduation rate.According to past teaching experience,student achievement determines the student’s pass rate to a certain extent.In this paper,the study of student pass rate is mainly based on the student’s course pass rate and the student’s graduation pass rate.Through literature reading and previous teaching research experience,it is found that student pass rate is affected by many factors,such as parents’ educational background,whether students get additional teaching support,etc.Both teaching managers and individual students hope to identify important factors that influence course pass rates and graduation rates.For the teaching administrators,to find out important factors affecting the student passing rate in time,we can propose a targeted teaching plan,change the teaching policy in time,truly realize the personalized guidance to the students,improve the student passing rate,and thus improve the teaching quality.For individual students,timely understanding of their current learning status can be targeted to change their knowledge blind spots,enhance their interest in learning,and avoid negative emotions(dropouts,schooling,etc.).Therefore,it is important to find out the influencing factors by predicting the student pass rate.In the past,the administrators of education and teaching roughly guessed the passing rate of students based on the existing teaching results.However,this does not meet the needs of education and teaching managers.Managers hope that in some way,the student pass rate can be predicted scientifically and accurately,instead of guessing the next result based on the existing results.Therefore,it is an urgent problem to be solved in the field of education to find important factors affecting student pass rate and predict student pass rate.In recent years,machine learning technology has performed well in all walks of life and has been widelyrecognized and promoted,especially in the field of education.Researchers have used machine learning algorithms to explore both student pass rates and key features of students.At present,the education model mainly includes two types: online education and offline education.In online education,researchers have been involved in the field of student pass rate prediction,and building predictive models has become a research hotspot.At present,the deep neural network algorithm has not been introduced in the online education research,and there is no consideration of whether there is a dependency relationship among students.There are characteristics of the graphic structure and the influence of the initial value of the coefficient on the algorithm.In offline education,the decision tree algorithm is simple and easy to implement;the support vector machine has advantages in the two classification problem.Researchers have used decision tree algorithms and support vector machine algorithms to explore the rate of student pass rate prediction.However,the current research also considers the existence of dependencies between features or the generation of graphical structures between features,the initialization values of coefficients and parameters and other issues affecting the algorithm.In summary,the main work done in this paper can be summarized as follows:(1)For online education,build an improved deep neural network prediction model(FDNN).Considering dependence of student characteristics and the influence of the characteristics of the existing graphical structure on the algorithm,a new concept of feature dependence and DAG method is proposed,and the feature dependence and DAG method are applied to optimize the deep neural network algorithm to improve prediction accuracy;considering the influence of the initial coefficient on the algorithm,an initialization coefficient rule is proposed to make the algorithm converge as soon as possible.(2)For the offline education,a prediction model based on the improved decision tree algorithm(FGDT)and a prediction model based on the improved support vector machine algorithm(FGSM)are constructed.Decision tree algorithms and support vector machine algorithms have shown advantages in the field of education,However,the current research does not consider whether there are dependencies between features and the characteristics of the existing graphical structure,the initial value of the coefficients,and the influence of the parameter problem on the algorithm.Aiming at these problems,this paper proposes that feature dependence and DAG methods can solve the influence of student characteristics and existing graphical features on the algorithm results,so as to improve the accuracy of the algorithm.This paper proposes the effect of the initialization coefficient rule on the algorithm results.In order to make the algorithm converge as soon as possible;this paper introduces grid search algorithm optimization decision tree algorithm and support vector machine algorithm to improve the accuracy of the algorithm.(3)Experiment analysis.This article conducts experiments on two modes of online education and offline education.In online education,the FDNN algorithm is compared with the decision tree algorithm and the support vector machine algorithm to predict the student pass rate.The experimental results show that the FDNN algorithm has better performance and better performance.In online education,FGDT algorithm and FGSM algorithm are compared with decision tree algorithm and support direction respectively.Comparisons of algorithms are made to predict student passing rates and find out the important characteristics of students that affect the passing rate of students.The experimental results show that the FDNN algorithm has better effect and better performance.