Modeling retail establishments’ freight trip generation: a comparison of methodologies to predict total weekly deliveries

Modeling retail establishments’ freight trip generation: a comparison of methodologies to predict total weekly deliveries

0.00 Avg rating0 Votes
Article ID: iaor20173671
Volume: 44
Issue: 5
Start Page Number: 1195
End Page Number: 1212
Publication Date: Sep 2017
Journal: Transportation
Authors: ,
Keywords: transportation: general, transportation: road, retailing, simulation, vehicle routing & scheduling, planning, statistics: regression
Abstract:

Assuming freight trip generation as the total number of freight vehicles arriving to retail establishments, for loading/unloading purposes and within a defined time period, we experiment and compare four alternative modeling methodologies to predict freight trip generation. The aim is to achieve better freight trip generation models, thus contributing to improving the chances of correctly dimension, for example, the parking infrastructure required to accommodate demand, or estimating the freight traffic impacts at micro level. Representing the state of the practice, the first two methodologies are based on cross‐classification/category analysis. The third methodology uses a generalized linear model specification, a robust alternative to ordinary least squares linear regression. The fourth methodology consists in the exploration of a dependent variable simplification using an Ordinal Logit model. The main source of data is an Establishment‐based Freight Survey, which collected data from 604 retail establishments in the city of Lisbon, Portugal. The selected independent variables were the establishments’ industry category, number of employees and retail area. The analysis allowed for the conclusion that (a) variable contribution varies depending on the chosen modelling methodology, (b) there is little variability in the quality of predictions depending on the selected model, but a considerable improvement in correct predictions can be achieved by reducing the variability of the dependent variable, and (c) the proposed indicator framework is suitable to compare model predictions and might be adequately represented by subset of those applied.

Reviews

Required fields are marked *. Your email address will not be published.