Cellular automata (CA) have been increasingly used to simulate complex geographical phenomena. These models may have limitations on reflecting individuals' behaviors which should be considered in urban simulation. Agent-based modeling can solve some of the problems in addressing individuals' influences in urban systems. However, there is a general lack of methodology on how to define agents' properties. This paper uses multicriteria evaluation techniques to determine some of the parameters for the agent-based model. Empirical data from GIS are used to define agent's properties. Sensitivity analysis is also carried out to assess the influences of parameters on simulation outcomes. This model has been applied to the simulation of the residential development in a fast growing city, Guangzhou, in south China.