Baseline Predictive Analysis of Saudi Digital Payment Activity Using Official Payment and Public-Sector Data

Saudi Arabia’s expansion of digital payment activity creates an opportunity for empirical analysis based on official payment systems and public-sector data. Existing Saudi studies have documented fintech development, digital-payment adoption, open government data, and COVID-related payment shifts, but less work has applied baseline predictive analytics directly to official Saudi payment and contextual datasets. This paper develops a modest applied benchmark using a region-month structure, with the main target variable being the monthly point-of-sale (POS) transaction value by region. The study compares payment-only models, based on lagged POS value, ATM cash-related controls, payment variables, moving averages, growth rates, and seasonal indicators, with payment-plus-contextual-public-data models that additionally incorporate CPI or inflation, population, business establishments, tourism activity, regional indicators, and calendar effects. Standard predictive metrics, including RMSE, MAE, MAPE, and R 2 , are used to evaluate temporal-holdout performance. The strongest results are achieved by LightGBM with contextual public-sector variables, while payment-only models also establish a strong empirical baseline. SHAP values are used only to support the interpretation of fitted tree-based models. The contribution is therefore intentionally limited but useful: it shows that official Saudi payment and public-sector data can support a clean, reproducible baseline predictive analysis without making causal, real-time, or high-claim explainable-AI assertions.