Integrating machine learning and meta heuristics in protecting data privacy and statistical properties

This study offers an innovative approach to preserving the privacy of sensitive data types, such as health data. Considering the tendency of traditional methods to reduce data analysis accuracy, the integration of machine learning and metaheuristic methods is aimed at achieving both data privacy and the preservation of statistical properties without compromise. In this study, experiments were conducted on synthetic patient data using genetic algorithms and differential privacy mechanisms. Initially, privacy levels were optimized through different epsilon values, followed by data analysis using classification algorithms. Additionally, Principal Component Analysis (PCA) was employed to reduce data dimensions, and the results were evaluated. The findings demonstrate that the Laplace mechanism effectively ensures privacy for both continuous and categorical data, and genetic algorithm-optimized parameters play a significant role in balancing accuracy and privacy. The XGBoost algorithm achieved the highest accuracy scores in the integrated datasets. This study provides significant contributions to the literature by developing more reliable and efficient analysis methods for sensitive data types through the combination of differential privacy methods and metaheuristic algorithms.