Classification of Student Graduation Using Backpropagation Neural Network with Features Selection and Dimensions Reduction

Students' graduation is a very important thing in universities as a determinant of the quality of a university. To improve the number of student graduates, a university must know the factors that affect students and classify student graduation. The efforts that can be made to detect student graduation are by predicting student graduation early on using data mining. The data mining technique used is using feature selection techniques with Mutual Information and Chi-Square, then dimensional reduction with PCA which will later be classified using BPNN. The dataset in this study is data on graduates of the Information Systems Study Program for the 2013-2022 academic year which contains 7 attributes with 38 questions. The 7 attributes are GPA, Total Credits, Taking FP Course on 7th Semester, Number of Repetitive Courses, Procrastination, Confidence, and Disciplin. The best experimental results can be obtained in BPNN classification with Mutual Information feature selections with 100% accuracy and RMSE 0 at a learning rate of 0.005 with a maximum iteration of 1000. While the results of the attributes that greatly affect student graduation are procrastination, discipline, confidence, and total credits.