Until now, academic data only used for the purposes of calculating GPA, graduation requirements, transcripts, and others processes related to reports and regular information. Academic data might be a tremendous source of information on educational improvement. One of the information which can be exploited is information on the predictive value of student in the future period (e.g. the next semester). This study attempted to use a data mining approach to help the guardianship process. The guardianship only needed the estimated range of GPA value to be able to determine whether the student is in the safe position or not based on academic rules and regulations. Data pre-processing implemented feature selection, raw selection, discretization and data sampling to get the best result. This study implemented Naive Bayes Classifier (NBC) and simple linear regression as the prediction method. The result from both methods will be compared at last. Comparison results from NBC and simple linear regression in several scenarios shown that NBC with data sampling got the best accuracy among other scenarios. The result means that data mining approaches is worth enough to be considered as a solution for predicting student's academic success.