Abstract—Every year, all the colleges hold new student
enrollment. It is needed to start a new school academic year.
Unfortunately, the number of students who resigned is
considerably high to reach 837 students and caused 324 empty
seats. The college’s stakeholders can minimize the resignation
number if the selection phase of new students is done accurately.
Making a machine learning-based model can be the answer. The
model will help predict which candidates who potentially complete
the enrollment process. By knowing it in the first place will help
the management in the selection process. This prediction is based
on historical data. Data is processed and used to train the model
using the Adaboost algorithm. The performance comparison
between Adaboost and Decision Tree model is performed to find
the best model. To achieve the maximum performance of the
model, feature selection is performed using chi-square calculation.
The results of this research show that the performance of Decision
Tree is lower than the performance of the Adaboost algorithm.
The Adaboost model has f-measure score of 90.9%, precision
83.7%, and recall 99.5%. The process of analyzing the data
distribution of prospective new students was also conducted. The
results were obtained if prospective students who tended to finish
the enrollment process had the following characteristics:
graduated from an Islamic school, 19-21 years old, parents'
income was IDR 1,000,000 to IDR. 5,000,000, and through the
SBMPTN program.