Data Analysis Of College Practice Teaching Based On Rotation Forests And LightGBM Classification Algorithm

Undergraduate practical teaching is the key link to improve the quality of talent training for college students.In recent years,major universities in China are making every effort to develop practical teaching construction for undergraduates.In this environment,Jilin University has established an undergraduate practical teaching system that meets the school's conditions,and began to officially start the construction of a practical teaching information platform in 2013.After years of practical operation,the practical teaching management platform has collected a large amount of real data of students in practical teaching.A large number of researchers have begun to carry out data mining research on the potential information in such big data in education.Generally,in the research of educational data,the main methods used include data mining methods such as classification,clustering,and association rule mining.In this paper,a new classification model is used to conduct data mining research on the practical teaching management data set.There are many types of classification algorithms,but due to the complexity and diversity of practical applications,the most widely used classification algorithm is an integrated classification model composed of multiple single classifiers.In recent years,the leader in the classification of machine learning competitions is the LightGBM algorithm developed by Microsoft.And because the rotation forests algorithm excels in the classification of small and medium data sets,this paper proposes an ROF-LGB model based on the integrated classification model of rotation forests and LightGBM,in an attempt to deepen the potential information of the practical teaching management data set.In this paper,a large number of comparative experiments are performed on different public data sets using the ROF-LGB model and other various treeclassification models.The experimental results prove that the ROF-LGB model has advantages over other classification models in accuracy and stability;In comparison experiments on practical teaching data sets,the accuracy and stability of the ROF-LGB model also exceeds that of the LightGBM model and the linear support vector machine model that excels in small data sets.In this paper,based on the feature importance score attributes in the ROF-LGB model,a feature importance analysis is performed on the student data of the Software College of Jilin University.After multiple sets of repeated experiments with different data sizes,this article compares the results of each set of experiments,analyzes the distribution of the importance scores of all features,and finally draws the following conclusions: 1.For students of the Software College,the basic information age,birthplace,and gender do have a greater impact on students' graduation development direction.The nationality,class,and absence of subjects have little effect on the students' development direction.2.For the students of the Software College,the most important thing is their own grades.Secondly,participating in the Discipline Competition,participating in the College Students Innovation and Entrepreneurship Training Project,participating in the Optional Experiment,applying for software copyrights,and participating in the Open Innovation Experiment,these types of extracurricular practice projects also have a greater impact on the development of students.In addition,the College English Test is of great help to students' advancement and employment.3.For the practical teaching construction of Jilin University,the excellent results of the Discipline Competition,the College Students Innovation and Entrepreneurship Training Project,and the Optional Experiment proved that it played an important role in student development.However,the Open Innovation Experiment has a low score on the importance of student development,and further improvement is needed to increase its influence on students' future development.