Predicting Internet Usage for Digital Finance Services: Multitarget Classification Using Vector Generalized Additive Model with SMOTE-NC

Digital Finance Service has a prominent role in the digital economy. Digital economy can be interpreted as economic and business activities through markets based on digital technology or internet and web technology. Practically, the internet has many purposes not only for entertainment and communication but also for financial services. Therefore, based on demographic characteristics, such as education, occupation, gender, race, age, and place of residence, this study aims to predict internet usage for buying, selling, and banking facilities. This is a classification problem with imbalanced multitarget classification, then the classification method is vector generalized additive model (VGAM). Also, we used Synthetic Minority Over-sampling Technique Nominal-Category (SMOTE-NC) to handle the imbalanced case. The dataset used is derived from the National Socio-Economic Survey (NSES) in 2020. The sample of this research is household members residing in urban districts or villages located in the province of East Java. The result shows that VGAM SMOTE-NC produces a mean geometric accuracy value obtained is 93.1% and can predict the minority class.