There is a gap in the number of students in and out of graduating on this study program. The gap occurs due to the low graduation of students on time. Therefore, this study aims to design a model of student graduation predictions on time and not on time in finding solutions to that gap. The predictive model used in this study is Naïve Bayes. The data used in the form of 44 graduate data in 2020 is divided into two parts of the analysis stage with RapidMider, namely 38 graduates (training) and six data for testing. Our findings showed that the resulting research prediction model was excellent, with five data from six graduates matching the predictions in the first test, while one data was illegible. However, in the second test, six graduate data are exactly the same as modeling, whose accuracy shows a value of 100%.