SUMMARY
The national exam is a benchmark for assessment carried out in the country for the parent of Education, which is used as a prerequisite for student graduation. Before taking the national exam, students take a trial exam which is used as a benchmark before the actual national exam. The use of graduation data is not yet realistic and maximum. This is quite difficult for students who are not known to have passed. In order to predict the pass threshold, it is possible to utilize existing data, especially completion data, to predict the completion rate. Due to the large amount of data, it requires a period to predict, so we need a system that can extend the estimated period of the student's graduation phase. This study describes the testing of prediction applications for passing the National Examination while using the black box method. The black box method consists of several methods, including splitting equivalence, limit analysis, comparison testing, sampling testing, robustness testing, and others. Among these tests, the marginal analysis test method was chosen for this study. Marginal analysis is a test procedure that determines the basic and upper bounds of the data to be tested. This test is run using the class feature added to the National Examination Pass Prediction application. The results of this test indicate that there are many weaknesses in validating the data and the data stored in the database does not match the desired data. The results of this test can be used as suggestions or evidence for application improvement.