PREDICTING DIGITAL FRAUD RISK USING SUPPORT VECTOR MACHINE CLASSIFIER A CASE STUDY OF UNIVERSITAS TERBUKA STUDENTS

Abstract

As a higher education institution that operates through open and distance learning, Universitas Terbuka depends significantly on digital infrastructure to support both its academic activities and student financial transactions. Students, as active users of digital technology are also vulnerable to various forms of digital fraud, such as phishing, identity theft, and scams related to social media or financial applications. Early detection of the risk of digital fraud is essential, allowing institutions to implement preventive strategies and deliver more targeted education on digital security. This study aims to develop a machine learning predictive model to identify the potential risk of digital fraud among students at Universitas Terbuka. A supervised machine learning approach using multiclass Support Vector Machine (SVM) was applied to a dataset collected from UT students who were actively studying and had reported income or employment status. Relevant features were selected and pre-processed to train the SVM model for multi-class classification of digital fraud risk levels. The model demonstrated a good classification performance, indicating that machine learning can effectively support early detection systems for digital fraud in open and distance learning environments. This study contributes to the emerging literature on fraud detection in the education sector by applying a Support Vector Machine model tailored to the context of a large-scale distance learning university.