Transport Timeliness Evaluation Using Gaussian Process Regression in Smart City Transit Systems

Timeliness evaluation of transportation is a key element of smart cities transit, as it drives operational efficiency, passenger level of satisfaction, and urban mobility management. Timeliness evaluation must be supported by effectively predicting travel times to ensure better scheduling and enhance reliability of day-to-day public transportation. However, the conventional techniques used to predict travel time accuracy often suffer from poor flexibility to model real time travel time data, incapable of capturing complex non-linear ways movements occur and limited ability when faced with irregularly spaced and sparse data. To this end, this research developed a new framework called Travel Time Predictions using Gaussian Process Regression (TTP-GPR). In combination with the probabilistic and non-parametric nature of Gaussian Process Regression (GPR), it is possible to effectively model uncertainties and complex spatiotemporal structures found in travel data. The TTP-GPR framework incorporates relevant traffic features and considers temporal changing dynamics, effectively predicting all travel times across various urban contexts. Experimental results on real transit datasets show that TTP-GPR consistently outperforms conventional transport models across four critical parameters with top scores of 5 each. It delivers superior travel time accuracy under various conditions, enhances punctuality through precise real-time predictions, provides detailed and reliable delay distribution analysis, and offers robust modeling of dwell time variability all with explicit uncertainty quantification for every travel time estimate.

Hali tarjima qilinmagan