THE ANALYSIS AND APPLICATION OF CLASSIFICATION METHODS FOR DEVELOPING SOFTWARE INTENDED FOR STUDENT ADMISSION BASED ON APTITUDE AND INTEREST SCOUTING SISTEM (CASE STUDY AT STT TELKOM)

The admission process for university students through aptitude and interest
scouting can be done by predicting the possible GPA that students may achieve.
This can be done by using data mining classification methods. The classifier is
developed based the student’s historical data related to their achievement in their
high school, the rank of their school and the major they choose.
The classification methods further analyzed was decision tree and Bayesian
network. These methods are often used in the solving problems in the area of
prediction and their level of complexity and accuracy are similar. Specially related
to decision tree, method of C4.5 is chosen because it has been widely used.
Meanwhile, Naïve Bayes was chosen because the assumptions within this method
are regarded to be more precise than those other of methods.


Based on accuracy level, the result shows that there is no absolute superior
method. The accuracy level of these methods has not been satisfying. In term of
their process time, C4.5 needs shorter time than Naïve Bayes. However, the
process time for both methods are applicable. For C4.5, there is no correlation
between data dispersion and accuracy levels. For Naïve Bayes, it can be found
that there is dependency between the attributes. Thus, Naïve Bayes is not suitable
to process the data of this experiment. For the purpose software implementation,
the level of accuracy of the methods needs to be increased.