Research On Student Learning Behavior Analysis Based On Xgboost Algorithm

With the development of education informatization,various learning behaviors of students generate a large amount of data.To make data work,using machine learning algorithms to explore and utilize the value of educational data has become the focus of research scholars.The emergence of massive data provides a lot of potential value for educational administrators,helps to improve the management efficiency of managers,education administrators can make overall planning based on student data information,starting from teaching mode management and teaching mode management,improve management level,provide early warning for those at academic risk.According to the specific situation of the graduates’ destination,suggestions on the specific implementation of their learning behavior for different graduation destinations such as postgraduate entrance examination,employment,entrepreneurship.The main work done in this paper is as follows: One is to collect data on students’ learning behavior,clean and integrate the undergraduate grades table and the graduate destination table,conduct classification statistics and analysis,extract various desired features,classify positive and negative samples,form three tables:high-quality graduate table,determination table,postgraduate employment table.The second is to use the xgboost algorithm,the knn algorithm and the Bayesian algorithm on the three tables of the high-quality graduate table,the determination of the destination table,the postgraduate employment table,respectively,to compare the performance evaluation results of accuracy,precision and recall,and make a three-dimensional clustered column chart to visually display the contrast effect.The third is to use the grid search method to adjust the optimal parameters of the xgboost algorithm model on the three tables of the high-quality graduates table,the destination table,the postgraduate employment table,to improve the overall performance of the model,and get the feature importance score ranking of the three tables,get the factors that affect where students graduate,get the accuracy,precision and recall of the xgboost algorithm modeled on these three tables and predict where the graduates go.In this paper,the xgboost model algorithm is used to model the three tables: the high-quality graduate table,the determination table,and the postgraduate employment table.The obtained performance evaluation of accuracy,precision and recall is generally better than knn model algorithm and Bayesian model algorithm.The method of adjusting the optimal parameters of the xgboost algorithm based on grid search can improve the accuracy,precision and recall of the model.According to the feature importance score attribute of the xgboost algorithm,this paper compares the experimental results of each group,analyzes the importance score of each group,and finally draws the following conclusions:1.The results of the college entrance examination,public compulsory courses,professional compulsory courses,and the first grade 4 have an important impact on the future of graduates.2.Age has little influence on graduation destination,and the relationship between graduation destination and age is weak.3.The accuracy rate of graduates’ destination prediction is about 70%,and some recall rates reach 90%.According to graduates’ grades and other learning behavior data,it can provide important reference for the prediction of graduation destination.