论文部分内容阅读
This paper aims to identify the relationship of the evolutionary elements and their solution mechanisms in genetic algorithms with respect of zone design problems.For zone design,the application of robust spatial modelling is essential to achieve the best solution so that the identification of search mechanism of genetic algorithms can be a good initiative for zone design studies,especially to develop a new solution part.For tackling this research objective,this paper develops p-median problem based algorithm in respect of practical level in zone design.Location modelling has a long history of practical application within geography,and zone design issues have been also widely applied to various human geography research themes.Hence,the location-allocation problem represents a hard optimisation problem,such as non-linearity in optimal solution procedures,they provide a good experimental ground for this research.To tackle the zone design application,this paper explores a spatial evolutionary algorithm with location-allocation model for which experimental criteria of evolutionary search mechanism are developed.Official geographies are no longer appropriate for general purpose analysis and mapping.On the one hand,arbitrary boundaries do not capture the nature and extend of the underlying socioeconomic phenomena.On the other hand,even when zones are designed for a specific purpose (such as census enumeration),it is not safe to assume that the same boundaries can be used for representing people and places for any other purpose.Zone Design has been suggested as a method for controlling the modifiable nature of areal units (Openshaw,1984).The method has been developed successfully for a number of different purposes and a whole range of scales.From designing census output geographies and areas for policy making (Alvanides & Openshaw,1999),to applying geocomputation for handling zones at the national level (Openshaw & Alvanides,1999),Zone Design proved to be an indispensable tool for geographical analysis.In simple terms,Zone Design involves the aggregation of small areas into larger zones in an automated manner,by optimising the spatial and statistical properties of these output zones.The optimisation method proceeds according to specific constraints imposed on the nature,number,shape and statistical properties of the output zones.The most important constraints being the internal contiguity of the output zones and the external function that needs to be optimised.Typical examples of functions involve equal population or physical area zones and optimised location-allocation models.More complicated models have also been applied so that output zones meet predefined correlation coefficients,or statistical measures applied in a multi-variate context (Alvanides & Openshaw,1999).The approach adopted in this paper involves the redesign of Output Areas (OAs) to Dong-like zones that optimise certain criteria.Dong is the smallest administrative boundary level and OAs is the second finest level of census output in terms of area size,but their variable size and high detail prevents their use without further manipulation.The OAs offer an ideal opportunity for Zone Design by user-defined criteria.In this case,the function sought to be optimized is a traditional location-allocation model applied on the OAs for the whole Seoul in order to produce zones with minimized travel cost.