Investigation of classification algorithms of machine learning in the field of education

With this study, we tried to determine which of the machine learning algorithms can be used in the field to form the basis of big data studies in education. Accordingly, in the study, factors consisting of 28 questions and five features, which are thought to be related to the determination of the performance of instructors in order to investigate the real data set and the quality of instructor, the Student Assessment of Turkey, are: The Decision Tree was examined under the algorithms of Random Forest and Artificial Neural Networks. The classification performances of these three algorithms were researched with accuracy, precision, specificity, precision, and F criteria. The study's data are gathered from the University of California's Machine Learning Data Repository. The data set consists of evaluation of three different lecturers with an likert-type scale by 5820 Gazi University students. R software and R Studio media were used to analyze the data. As a result of the study, the decision trees algorithm was found to be more successful than the Random Forest and Artificial Neural Networks, with an accurate classification ratio of 98.57%. The Random Forest algorithm was identified as the second best performing algorithm with a value very close to the decision tree at 98.03%. The Artificial Neural Networks performed at an acceptable level (81.55%) but are considered lowperformance compared to other algorithms. Again, the Decision Tree algorithm was found to be more successful compared to Random Forest and Artificial Neural Networks when analyzed by using criteria like specificity, sensitivity ratio, accuracy ratio and F. The study investigated a comparison of the most significant predictor and materiality levels of three algorithms as the sub-problem of the study. In conclusion, the most important predictive class variable for the three algorithms was found, and no difference in level of magnetism was encountered among the three algorithms.ass variable and no difference was ecountered in terms of the level of magnetism.