Proceedings of ICLT 2024
COLLABORATIVE LEARNING FOR DEMAND FORECASTING IN URBAN LOGISTICS
Maharshi Dhada; Duncan McFarlane
Institute for Manufacturing, Department of Engineering, University of Cambridge, Cambridge, UK; Institute for Manufacturing, Department of Engineering, University of Cambridge, Cambridge, UK
International Conference on Logistics & Transport 2024, Seoul, South Korea, pp. 31-36
Download PDF | View interactive page
Abstract
Purpose: This paper presents a probabilistic hierarchical statistical model for tackling the often encountered cold-start problem in urban logistics sector. Companies face the cold-start problems when they start their operations in a new city and therefore don't have sufficient data to accurately forecast the demand for that city. Using real-world industrial data, authors propose a hierarchical autoregression model that enables a logistics company to forecast customer orders in a new city, given the historical orders in other cities. Design/ methodology/ approach: Using a statistical hierarchical model enables the user to systematically model the operations of an urban logistics company across multiple cities. Each city is associated with an individual autoregression model, whose parameters are sampled from a common higher level distribution that represents the general behaviour of other similar cities. Operations in a new cities are therefore enhanced using systematically evaluated prior knowledge, enabling reliable forecasts in the early time-steps. Similarities across the cities are modelled using factors such as population density, geography, cultural influences, etc. Findings: Using a statistical hierarchical model enables the user to systematically model the operations of an urban logistics company across multiple cities. Each city is associated with an individual autoregression model, whose parameters are sampled from a common higher level distribution that represents the general behaviour of other similar cities. Operations in a new cities are therefore enhanced using systematically evaluated prior knowledge, enabling reliable forecasts in the early time-steps. Similarities across the cities are modelled using factors such as population density, geography, cultural influences, etc. Originality/ value: This is the first application of statistical hierarchical modelling for forecasting customer demand in urban logistics sector. Such a cold-start problem addres
Keywords
Machine Learning; Logistics; Collaborative Learning; Statistics; Hierarchical Modelling
Citation
Maharshi Dhada; Duncan McFarlane (2024). COLLABORATIVE LEARNING FOR DEMAND FORECASTING IN URBAN LOGISTICS. Proceedings of the International Conference on Logistics & Transport (ICLT 2024), Seoul, South Korea, pp. 31-36.