ASSESSING THE ROLE OF STATISTICAL MODELING TECHNIQUES IN FRAUD DETECTION ACROSS PROCUREMENT AND INTERNATIONAL TRADE SYSTEMS

Abstract

This study addresses the problem that fraud detection in procurement and international trade workflows often depends on rule-based checks and fragmented records, weakening detection and investigation prioritization. The purpose was to quantify how statistical modeling techniques relate to fraud detection effectiveness in enterprise case environments. Using a quantitative cross sectional, case-based design, 220 usable survey responses were retained after screening (12 removed from 232; item missingness 1.6%), covering procurement (n=108, 49.1%), international trade (n=92, 41.8%), and hybrid exposure (n=20, 9.1%). Key variables were descriptive analytics use (DAU), correlation-based screening (CBS), regression modeling practice (RMP), data readiness (DR), process control context (PCC), and fraud detection effectiveness (FDE), measured as 1–5 Likert composites with strong internal consistency (α=.81–.88). The analysis plan applied descriptive statistics, reliability testing, Pearson correlations, and multiple regression predicting FDE from DAU, CBS, RMP, DR, and PCC with controls. Respondents reported moderate to high adoption (DAU M=3.92, CBS M=3.71, RMP M=3.54) and relatively high effectiveness (FDE M=3.84). FDE correlated with DAU (r=.52), CBS (r=.45), and RMP (r=.49), all p<.001. In regression, the model was significant (R²=.48; F (7,212) =27.61, p<.001) and showed that DAU (β=.24, p<.001), CBS (β=.15, p=.017), RMP (β=.21, p=.001), DR (β=.18, p=.003), and PCC (β=.17, p=.004) each contributed uniquely to FDE. Subgroup models indicated procurement emphasized descriptive monitoring and controls (R²=.51) while trade emphasized regression scoring and data readiness (R²=.44). Implications are that fraud programs should implement layered, interpretable analytics and invest in data readiness and process controls to translate modeling capability into detection gains.