Academic Early Warning Research Based On Machine Learning

In the process of constructing a smart campus in colleges and universities,academic early warning is the main component of the smart campus.It mainly uses data mining technology to ensure that students complete their studies successfully,and at the same time provide certain decision-making support for colleges and universities.The thesis aims to build an academic early warning system to ensure that students complete their studies smoothly.The main research content is divided into two parts: "Academic Early Warning Model Research" and "Academic Early Warning System Design and Implementation".The details are as follows:In the process of studying the early warning model of academic work,based on the historical performance data,all-in-one card consumption data,and library card-swiping record data of undergraduates in a certain school,it is predicted whether they will be repetitive after the end of the fourth semester.First,a heuristic missing value filling method is proposed in the data preprocessing to obtain data features that can be used for repeating prediction.Then analyze the correlation between different features and grade repetition,perform feature screening and form different feature sets,respectively use random forest,decision tree,support vector machine and logistic regression four commonly used machine learning classification algorithms to construct academic early warning models,and comprehensively evaluate Then choose the early warning model with the best predictive effect.Finally,in order to solve the over-fitting problem of the selected model in the prediction process and improve the prediction effect of the model,a model based on genetic algorithm improvement and a model combination optimization method based on voting integration are proposed.The model with improved algorithm has the best evaluation effect,with Recall reaching 89.35%,Precision reaching 80.96%,and AUC reaching 84.25%.During the design and implementation of the academic early warning system,based on the software requirements of the relevant departments of a school,on the basis of the academic early warning model,the Zhima PHP framework,the PHP programming language and the Python programming language,and the Mysql database were used to achieve academic early warning.system.The system realizes the performance query and data export,and can provide academic early warning from the data of multiple dimensions of students,to ensure that students complete their studies normally,and provide certain decision-making support for colleges and universities.