Agent-based Household Micro-Datasets: An Estimation Method Composed of Generalized Attributes with Probabilistic Distributions from Sample Data and Available Control Totals by Attribute

Abstract

The purpose of this study is to build a system to rationally estimate the agent-based household micro-dataset of the base year for land-use microsimulation. Attributes of a household can be classified into a general set of categories. This system, wherein a Monte Carlo simulation is used, deals with those attributes, either continuous or discrete, in a generalized scheme. It uses sample data that contain full information on the micro-data to establish the correlation between the attributes and the available statistical data or census data. To reproduce the correlation between continuous attribute variables, independent variables that can be obtained based on the sample data are introduced and employed as intervening variables. Attributes of a whole household are probabilistically determined based on Logit and other models obtained from sample data. Finally, a case study of the system application to a person-trip-survey dataset of the Sapporo metropolitan area is presented.