A Quantitative Study on the Impact of National Demographic Characteristics on Cybercrime Distribution Based on Spearman and XGBoost Models

Abstract

This study focuses on quantitatively analyzing the relationship between national demographic characteristics and the distribution of cybercrime, aiming to identify the most influential macro-factors. The research first employed Spearman's rank correlation analysis to assess the linear correlations between twelve national indicators—encompassing economic, social, educational, and internet usage metrics—and the cybercrime factor. The results revealed strong linear correlations for GDP per capita, GDP, aging rate, and three internet-related data points. Subsequently, to incorporate both linear and non-linear relationships, this study utilized the XGBoost machine learning model for in-depth analysis.Feature importance analysis from the XGBoost model indicated that GDP per capita is the most influential national characteristic affecting cybercrime distribution, with a correlation coefficient of 0.24633. The number of secure internet servers and GDP were also identified as high-impact indicators. The findings suggest that while regions with high GDP per capita host active online financial activities and advanced technology—potentially offering more criminal opportunities—overall, economically developed areas may exhibit lower rates of cybercrime incidence. In contrast, the unemployment rate demonstrated the lowest impact, with a coefficient of only 0.01276. The model achieved a mean squared error (MSE) of 0.339 and a coefficient of determination (R²) of 0.744, demonstrating strong explanatory power and predictive accuracy regarding the factors influencing cybercrime distribution.