Predictive Justice in Data Privacy: Explaining GDPR Fine Determinants Through Legal Feature Engineering and Ensemble Machine Learning Models

The General Data Protection Regulation (GDPR) establishes a harmonized legal framework for personal data protection across the European Union. However, the substantial variation in administrative fines imposed by different supervisory authorities raises questions regarding the consistency and predictability of GDPR enforcement. This study aims to develop an explainable predictive justice framework for estimating GDPR administrative fines using legal feature engineering and ensemble machine learning while identifying the legal and institutional factors that most strongly influence enforcement outcomes. The proposed framework utilizes a dataset of 437 GDPR enforcement cases and applies logarithmic transformation to address the highly skewed distribution of administrative fines. Legal and contextual information are transformed into predictive features through one hot encoding and feature engineering, including violated GDPR articles, supervisory authorities, countries, violation types, case complexity, and the number of violated articles. The predictive performance of a Random Forest ensemble model is compared with a Linear Regression baseline using the coefficient of determination (R²), Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE). Experimental results demonstrate that the proposed Random Forest model significantly outperforms the baseline model, achieving an R² value of 0.800, compared with 0.530 for Linear Regression, with an MAE of 0.8545 and an RMSE of 1.1750. Feature importance analysis reveals that Summary Length, Supervisory Authority, Country, and Violation Type are the most influential predictors, whereas individual GDPR articles contribute comparatively smaller importance values. These findings indicate that administrative fine determination is influenced not only by statutory provisions but also by institutional and contextual characteristics of enforcement cases. The proposed framework contributes to the field of cyberlaw by demonstrating that explainable ensemble machine learning can effectively support predictive justice analysis while providing interpretable empirical evidence regarding regulatory consistency and legal decision making under the GDPR.