Enhancing Trust and Transparency in Metaverse Financial Systems Through Explainable Artificial Intelligence for Risk Assessment

This study proposes an Explainable Artificial Intelligence (XAI) model for financial risk assessment in the Metaverse ecosystem by combining predictive accuracy with interpretability through the XGBoost algorithm. The model was trained on behavioral, transactional, and demographic data to capture complex relationships influencing user financial risk. The evaluation results showed strong predictive performance, with an R² value of 0.813 and a 5-fold cross-validation R² of 0.816, indicating robustness and generalization. Feature importance analysis identified High-Value Purchase Pattern and New Users as the most significant predictors, followed by Login Frequency and Transaction Amount, highlighting the importance of user activity and experience in determining financial risk. Residual diagnostics confirmed that the prediction errors were normally distributed and unbiased, demonstrating that the model was accurate and fair across different risk levels. The integration of explainability mechanisms allows stakeholders to interpret and validate AI-driven decisions, promoting transparency and accountability. This research contributes to the advancement of trustworthy and ethical AI systems in virtual economies, offering a practical framework for transparent financial risk management within the Metaverse.