Akademik Veri Yönetim Sistemi
JOURNAL OF SUPERCOMPUTING, cilt.81, sa.4, 2025 (SCI-Expanded)
Intelligent transportation systems (ITS) involve collecting, storing, and delivering real-time traffic data to optimize efficiency, enhance transportation safety, and reduce energy consumption by integrating advanced electronics, information systems, and telecommunication technologies into roads, vehicles, and related infrastructure. These systems typically use some indicators such as speed, flow, and density to predict traffic information, which are measured by geographically fixed-position sensors located on the roads. Privacy leakage is a major problem since traffic information derived from users' or vehicles' mobility data contains sensitive information. Consequently, existing approaches are constrained by geographical dependencies, restrictions and privacy vulnerabilities. This study introduces a state-of-the-art, privacy-preserving, feature engineering-based, geo-independent traffic speed prediction framework to address these challenges for the first time. The framework, evaluated using performance metrics such as mean absolute error (MAE), mean absolute percentage error (MAPE), mean square error (MSE), and root-mean-square error (RMSE) on a real-world dataset, was implemented using long short-term memory, bidirectional long short-term memory, gated recurrent unit (GRU), linear regression, support vector regression, multilayer perceptron and random forest regression, and differential privacy with varying epsilon values. According to the results, GRU yielded the most favorable outcome with MAE of 7.7, MAPE of 28.5%, MSE of 132.7, and RMSE of 11.5 for Istanbul, MAE of 9.7, MAPE of 17.2%, MSE of 177.4, and RMSE of 13.2 for Eskisehir, and MAE of 3.1, MAPE of 5.4%, MSE of 12.4, and RMSE of 3.4 for Konya Roads with the epsilon value of 200, respectively. These findings highlight the framework's potential to advance privacy-aware traffic management and urban mobility solutions, offering a transformative approach for ITS applications.