The Multi-Objective Spatial Optimal Location Based On Genetic Algorithm

Spatial optimal location is one of the classic problems in the realm of GIS. It selected the most appropriate location for site or sites in a certain geographic area and made performance or a combination of performances to reach optimum. This problem usually involves high-dimensional geographical space, massive data, multiple conflicting objectives and constraints and the issue is a NP-hard combinatorial optimization problems. Thus, traditional methods such as brute-force, blindness search and etc. are infeasible to find the optimal solution. This paper demonstrates the method integration of the knowledge of spatial structure and association patterns into multi-objective genetic algorithm to executive spatial optimal decision based on GIS's powerful spatial data processing, spatial query, spatial analysis and visualization and so on technology.Firstly, the paper taking two basic aspects—qualitative and quantitative factors that affect the process of site location into account. Analysis, quantification and modeling of uncertainty qualitative factors, and deterministic quantitative factors analysis and modeling respectively, then quantified qualitative factors with a comprehensive utility factor using rough sets processing method and introduced the comprehensive utility factor into mathematical model of quantitative factors so as to realize the modeling processSecondly, based on the analysis of principle and internal processes of multi-objective genetic algorithm, we established the three models about multi-objective spatial optimal location, that is to say NSGA-II, SPGA2 and PESA-II that integration of the knowledge of spatial structure and association patterns. Meanwhile, we reached the compacted seamless integration of GIS and application model. Finally, design and development GIS desktop application system based on ArcGIS Engine9.3 + c #.net and conduct the experiments about spatial optimal location of Beijing hospitals in large irregular area. Experimental results show that the knowledge of spatial structure and association patterns has great influence on spatial optimal location, models created in the paper can be convergent to Pareto optimality sets more effectively, the distribution of the optimal location is better and improved performance than pure multi-objective genetic algorithms while maintained stability and robustness of model.