Estimation of a loading/unloading parking duration model for commercial vehicles in urban areas in the Medellin metropolitan area using lasso regression models

Over recent decades, the delivery of goods and services in urban areas has faced mounting challenges, including insufficient parking availability for deliveries, rising demand driven by e-commerce, and negative externalities such as congestion, pollution, and traffic accidents. This study develops a parking-duration model for commercial vehicle loading and unloading activities using Lasso regression, based on 247 recorded truck operations in Medellín, Colombia. The analysis identifies key determinants of parking duration—including cargo weight, units per pallet, number of employees, and the use of manual and mechanical handling equipment, as well as assisted delivery tools. While duration (survival) models such as hazard or Weibull approaches are commonly applied to time-to-event problems, this study adopts Lasso regression for practical reasons. Specifically, Lasso offers simultaneous regularization and data-driven variable selection, making it well-suited to small, noisy datasets. This choice is based on methodological fit rather than demonstrated empirical superiority over survival models for this dataset. The results indicate that the use of mechanical equipment significantly reduces dwell times, whereas greater cargo weights and reliance on manual handling increase them. By uncovering the operational drivers of freight parking duration in a Latin American metropolitan area, this research makes both methodological and practical contributions. The findings support planning-level design of regulated freight loading zones, targeted equipment adoption programs, and differentiated curb regulations calibrated to observed drivers' dwell times; evaluating welfare-improving or system-optimal outcomes would require additional simulation and/or field evaluation beyond this paper. Future research should aim to expand the dataset, integrate additional categorical variables, and compare results across cities in the Global South and North to enhance generalizability and policy relevance.