LOTUS-PARK: Intelligent Real-Time Parking Assistance using Machine Learning in Urban Smart Mobility Environments

LOTUS-PARK, a novel metadata-driven machine learning framework for intelligent, real-time parking assistance in urban environments. Unlike most traditional systems that process images, LOTUS-PARK employs structured contextual data (temporal, spatial, environmental and vehicle related features) to derive predictions of parking occupancy and suggest the optimal parking locations. The ability of the proposed methodology to do multi-stage learning is evidence of its abilities by capturing both pattern recognition and temporal stability, in addition to its incorporation of a predictive GreenScore metric to model sustainability. Experiments performed on a custom generated urban parking dataset show that LOTUS-PARK outperforms baseline models (Logistic Regression and Random Forest) with 0.90 of accuracy, 0.88 of F1-score with a precision of 0.91. Moreover, the system has a strong capability of predicting eco-efficiency, as the GreenScore regression R² is 0.96 with an MSE of 0.002. Analysis of occupancy trends across variables such as days, hours, weather, and lighting conditions confirm the model's robustness and flexibility. LOTUS-PARK offers a sustainable and intelligent solution for optimizing smart mobility and urban parking with high predictive accuracy.