Student performance analysis with data mining in distance education synchronous, asynchronous and hybrid courses in the pandemic process

In these days, the rate of university students' access to education and training materials from the internet has increased considerably. There has been an increase in the data set that can be analyzed due to the use of the internet in education and the increase in access to course materials. One of these data sets (planned or urgently) is the homework, exam, project, performance, attendance grades and the like accumulated in the distance education systems of the universities that started the distance education process. In the new Corona Virus pandemic, with the recommendation of the Higher Education Council, universities continued their education by using Asynchronous, Synchronous and Hybrid methods remotely and even had to take the exams in the distance education system. In this research, data obtained from the Distance Education Application and Research Center System of Kayseri University was used. There are 8319 processed data within the scope of the research. Using the automatic modeling feature of the RapidMiner program used in Data Mining on these data, future predictions were made with default algorithms. In this study, Deep Learning, Naive Bayes, Gradient Boosted Trees, Logistic Regression algorithms, which give the best results among the default algorithms of RapidMiner program, were examined in depth to get the best result by changing their parameters. In addition, the k-Nearest Neighbor algorithm, which is not included in this automatic modeling, is also included in this study. It has been tried to obtain better results by making changes on the parameters of these 5 algorithms. The model established with Logistic Regression gave the best estimation result according to student success with 73.50%. The effect of all of the participation methods (Synchronous, Asynchronous and Hybrid) on student achievement was compared with the Confusion Matrix method and it was seen that the most reliable method was Hybrid. With this study, inferences were made about which of the methods of attending lectures in universities would be more reliable for students. Therefore, with the inferences made, it has been possible to predict which of the methods of participation in the course is more reliable in order to increase the level of student achievement for the next academic term.