DEEP LEARNING–DRIVEN RIDE-HAILING PRICE PREDICTION FOR INTELLIGENT TRANSPORTATION SYSTEMS

Abstract

Accurate forecasting of booking value and short-term ride demand is increasingly important for enhancing operational efficiency in urban ride-hailing systems. However, existing forecasting pipelines suffer from inconsistent model performance and difficulty in adapting to sudden behavioural shifts in urban mobility patterns. Additionally, most current approaches struggle to generalize across different regions and time intervals, leading to unstable and unreliable demand predictions. Conventional machine learning approaches such as Linear Regression, Random Forest, and Gradient Boosting have been widely applied to predict ride demand; however, their performance is often constrained by their limited ability to model temporal dependencies and rapidly evolving urban mobility patterns. Linear Regression fails to capture complex nonlinear relationships, Random Forest struggles with sequential correlations, and Gradient Boosting requires significant computational effort when handling high-dimensional and dynamic datasets. To overcome these limitations, this work proposes an enhanced forecasting framework based on a Recurrent Neural Network (RNN) integrated with a Greedy Tree–based optimization algorithm. The RNN component effectively models time-dependent fluctuations in booking value and ride demand, while the Greedy Tree module optimizes feature selection, reduces redundancy, and improves model interpretability. Comparative evaluation across multiple realworld urban analytics datasets demonstrates that the proposed hybrid architecture significantly outperforms existing traditional models in prediction accuracy, responsiveness to sudden demand variations, and computational efficiency. The results highlight the potential of combining deep sequential learning with greedy optimization to provide more robust, scalable, and actionable insights for intelligent transportation systems and urban mobility planning.