This paper develops a method based on iterative proportional fitting (IPF) for generating synthetic populations for applications of Albatross, which is a rule-based and activity-based model of travel demand. The Dutch Travel Survey (OVG) is used to provide sample data for populations to be synthesised by zone. The paper investigates applicability of IPF even under conditions where location or year of the sample differs from that of the target population. Results indicate that spatial discrepancies can create significant biases in the synthesised population, whereas temporal discrepancies do not present a problem within the range investigated here. Segmentation of the OVG sample is proposed as a way to reduce biases caused by spatial discrepancies. CHAID is used to find an optimal segmentation based on urban density and region variables for this purpose.
Samenvatting