Prediction of KIP Scholarship Eligibility at Universitas Almuslim Using an Explainable Artificial Intelligence–Based XGBoost Algorithm

Abstract

The selection process for Kartu Indonesia Pintar (KIP) scholarship recipients at the higher education level continues to face several challenges, including subjective assessment, limited transparency, and the suboptimal use of data-driven decision support systems. This study aims to develop a predictive model for KIP scholarship eligibility at Universitas Almuslim using the XGBoost algorithm integrated with an Explainable Artificial Intelligence (XAI) approach. The dataset employed in this study consists of synthetic data constructed based on official KIP selection parameters, encompassing economic, academic, social, and demographic aspects, thereby ensuring the confidentiality of student data. The research stages include data preprocessing, predictive modeling, policy-based validation, and analysis of prediction results. The XGBoost algorithm is utilized to generate eligibility predictions along with associated probability scores, which are subsequently evaluated to ensure alignment with scholarship selection principles and regulations. The simulation results demonstrate a clear separation between eligible and non-eligible students, with prediction probabilities predominantly concentrated at extreme values, indicating a high level of model confidence. Further analysis reveals that economic indicators and social affirmation variables exert a more dominant influence than academic factors, which function as supporting criteria. These findings indicate that the proposed system is capable of producing stable and consistent predictions while enhancing transparency and accountability in the decision-making process. This study proposes an interpretable scholarship eligibility prediction framework that can be adapted by other higher education institutions as a fair and data-driven decision support system.