improving delivery conditions by dynamically managing the urban parking system: parking availability prediction

Urban deliveries are a real challenge that the authorities often overcome by deploying a set of delivery areas. these static infrastructures are not compatible with either the highly punctual nature of the delivery activity or with urban dynamics. we propose an intelligent urban parking management system capable to modify in real time the status of any parking spaces, from a conventional place to a delivery bay and inversely. its driving force is the parking availability prediction. in this article, we model the number of available places using several types of regression combined by the stacking technique to improve the accuracy of predictions. we analyze their results, identify their strengths as well as their weaknesses and deduce the most effective one. we calculate the probability of a place being left free by the actuarial method. to do so, we exploit a massive amount of real-time parking availability data collected and disseminated by the city of melbourne, australia.