Application Of Data Mining In The Repetition Prewarning Of College Students

With the rapid development of information technology and the Internet industry,technology has successfully led the trend of the times.Big data technology is one of the most compelling developments.Everything speaks in terms of data,and everything is supported by data,creating today's big data trend.Nowadays,Big Data has extended to many fields and completed the integration of many resource.The development of science and technology has brought about tremendous changes in modern education.Electronic projection has replaced blackboard and online education has been widely accepted.How to properly use information technology to help education and the management of college students has become a research hotspot for scholars and government.Educational data mining technology is developed under this demand,and it effectively applies data mining methods to education,which has brought changes and innovations in the university.In educational data mining,predicting students' academic performance is a common research topic.It plays a very important role in the development and reform of college education,and has great practical significance and academic value.Recalling the development and reform of Chinese education in the past 40 years,the university organization and teaching management have been continuously improved,and the number of university student admissions has increased year by year,providing a large number of high-quality talents and scientific research elites for the country's development and innovation.However,in recent years,there has been an upward trend in the disciplinary suspensions and drop-out rate of college students.Early warning of high-risk students has become an important research area.Therefore,this paper conducted analysis and research based on the student's historical score data,card consumption data,and weblog data.Based on the records in the first three semester,we use various machine learning methods to train models and use these model to predict whether the student will be repeated in the end of the fourth semester.By preprocessing the student behavior data using methods like cleaning,desiccation and conversion,we calculated the statistical information of students' behavior data.After correlation analysis and feature screening,different combinations of features were used to build model and predict.The outcome is not bad,but the forecast methods need to be optimized.Through in-depth analysis of student performance,we use the K-Means clustering method to classify students' first four semester courses into five major categories based on the statistical information of the course.For courses that need to be predicted,we consider several algorithms like logistic regression,decision tree,naive Bayes,RNN and select the best one to build the prediction model.Combined with the failed credits of the previous three semesters,we analysis and predict the grade-reservation situation in the end of the fourth semester.Compared with the previous way,the effect is improved significantly.In order to further optimize the prediction process of the model,after understanding the training program and the curriculum arrangement,the unreasonable part of the course classification results in the experiment was intervened and optimized.Experiments show that classifying curriculum by the relevant characteristics of historical scores,and then modeling and analyzing students' fourth semester courses can effectively improve the forecasting results.The Recall of the optimal model reaches 100%,Precision reaches 74%,F1 value reaches 85%,which can achieve high precision in judging the students' risk of repetition.The high-risk student prediction method proposed in this paper combined with multi-dimensional data,not only applicable to a certain grade,but also can be well extended to different professions and colleges,which has strong practicability.At the same time,college leaders and counselors can help and intervene students according to the prediction results,reduce the repetition rate and drop-out rate of college students,and further improve teaching quality and student quality.