Cellular automata (CA) models are used a lot in urban planning for land use change simulation. Neighborhood rules in CA models are normally derived by analyzing raster maps, while in reality, urban planning is based on parcels. Moreover, new trends of land redevelopment require land use transition impacts to be addressed and incorporated in land use simulation. This can be accomplished by comparing neighborhood compositions of a particular site, before and after its transition. This paper presents a generic approach to semi-automatically discover neighborhood rules by analyzing vector maps, with a special focus on transition impact analysis. The approach contains one script for manipulating vector data and another script for visualizing neighborhood compositions. A step by step instruction of this approach is presented. The North Brabant region of the Netherlands is used as a case study area. Industrial site transition is used as an illustration for land use transition process. Three types of statistical comparison algorithms are used to compare the land use model which uses neighborhood rules discovered from the proposed approach with a benchmark model which only models land use self-influence. The robustness of the approach is studied by separating the region into urban and non-urban areas. Results show the applicability of this approach in improving transparency and applicability of CA models.