<![CDATA[Delays in tuition fee payments are a crucial problem commonly faced by private universities in Indonesia. This problem not only complicates campus financial management, but also has the potential to hinder the smooth running of student studies, such as leave of absence or discontinuation of studies. To date, there are not many predictive systems used to detect potential late payments early, especially the Random Forest (RF) model with Particle Swarm Optimisation (PSO). Therefore, this study aims to develop a predictive model for student payment delays by utilising the RF algorithm optimised using the PSO feature selection method. The Dataset used consists of 15,697 student data covering academic and administrative attributes. Pre-processing was carried out to convert categorical data into numerical form so that it could be processed by the classification algorithm. The evaluation results show that the RF model without optimisation produces an accuracy of 97.37%, precision of 100%, recall of 18.68%, and AUC of 0.825 ± 0.020. After feature selection with PSO, the model performance improved, with an accuracy of 98.83%, precision of 98.20%, recall of 25.40%, and AUC remaining stable at 0.825 ± 0.035. The most influential attributes in the classification were semester, leave status, studying while working, and father's occupation. The results of this study indicate that the combination of RF and PSO can produce an efficient and accurate prediction model, which can be used as a decision-making tool in higher education administration management.Keywords— Late Payments, Classification, PSO, Random Forest, Feature Selection.]]>
Copyrights © 2025
