Retrieving individuals´ attributes from aggregated dataset for urban micro-simulation: A primary exploration

As the traditional top-bottom based macro-simulation models cannot well adapt to spatiotemporally dynamic urban systems researches, the bottom-up micro-simulation models (e.g., multi-agent systems, MASs) based on individual data have gradually become a novel and dominating perspective for investigating urban dynamics. However, one of the factors restraining the wide application of micro-simulation models is l the poor availability of disaggregated individual data. Such situation is especially serious in China, because the individual data from the official census is not available and can only be obtained via various small-scale surveys conducted by separate units. The contribution of this paper is to propose a solution to retrieve individuals´ data from aggregated dataset, e.g. official statistical yearbook, census reports and contingent questionnaire reports, for urban micro-simulation under the current sparse-individual-data context. Based on the existing multi-source official statistical date, non-official surveys, as well as general relationships among individuals´ attributes (e.g., the marriage condition of a person depends on his/her age), this approach can inverse both the individuals´ attributions and spatial distribution. The approach was tested in the Beijing Metropolitan Area for retrieving the population spatial distribution and their attributes using the Fifth Population Census Report of Beijing, in which the aggregated information of populations is provided. The inversion results were proved to be significantly accurate in statistical level. Using the inversed samples´ as the input of the urban microsimulation models is promising to settle their bottle-neck problems for wide application.