| The simple information processing system of student performances that merely includes management,statistical analysis and comparison is no longer sufficient for the actual demands of students and colleges due to the rapid development of Internet technology and the increasing dependence on IT of college teachers and students.To this end,under the context of the extensive application of artificial intelligence into diversified fields,studies on the prediction of various student performances based on the data mining technology have currently become a popular topic both domestic and international in the academic area of college information management.Scholars have achieved different results on the prediction of various education-related performances through preprocessing the behavioral characteristics of all kinds of students with different data mining algorithms and from different perspectives.However many problems still require further researches,including which type of data mining algorithms to facilitate the development of college's information-based management,in what directions the Big Data technology will help colleges to manage students,how to apply the data mining algorithms to the student management and make it effectively,and how to predict the future performance of students with the existing data.In this paper,we raised some researches on the foundation of previous academic results and the current reality of informatization establishment in colleges,:Firstly,analyzed and compared different decision tree algorithms,verified the accuracy and applicability of them in the field of student performance by experiment.The experiment extracted and preprocessed the data of various characteristics such as the interests,habits,after-class reviews,attendances of students and the professional qualifications,teaching attitudes,teaching approaches of teachers to form a reasonable data collection,and then applied different decision tree algorithms to the prediction of student performances by using the 10-fold cross-validation to examine the effect and reliability of the algorithm.The result has demonstrated that the improved C4.5 algorithm generates higher accuracy than the original C4.5 algorithm and some other decision tree algorithms,as well as better robustness under training sets of different attribute dimensions.Secondly,based on the former part of the decision tree student performance prediction model,with the object-oriented thought,integrated the existing universities information system,we designed the system architecture,the system model and database structure,preliminary implementation a teaching evaluation system based on B/S architecture.The system achieve the interconnection and interworking between itself and other subsystems in the same college and thus to acquire various student information.The teaching quality evaluation and prediction system then predicts student performances through processing the existing information data.It is of great significance in guiding the teaching works and evaluations of colleges.Lastly,in the end of this paper,we also drew the conclusion of above major works,pointing out the deficiencies of this study caused by the lack of time and the objective problems in integrating other systems,namely the inadequate data choice of student performance related characteristics and the absence of academic exploration into the construction of teaching staff,the teaching profiles,the teaching conditions and other indicators that are also closely related to the teaching quality evaluation.In the future,I will continuously increase the attributes and functions covered by the system,and further involve the comprehensive application of the association rules algorithm,the decision tree algorithm and the clustering algorithm to the analysis towards the psychological statuses of students and the One-Card-Pass consumption and usage,in order to surgically assist and help students with the analytical results. |
PDF Full Text Request