摘要
由于案例推理接近于人类认识、解决问题最原始的思维方式，具有在无法获取机理模型、确定规则或统计模型时，采用简单的历史相似性实现问题的定量求解和预测的特点，成为当前人工智能中备受关注的领域。在讨论案例推理方法发展的基础上，探讨应用案例推理进行地学问题求解的具体方法。从地学数据分析的角度提出了地理案例推理，并结合地理案例的特点，具体讨论了地理案例的表达模型和推理模型，最后结合东海中心渔场预报的实际工作，给出这种方法的具体应用实例。
As an emerging branch of AI, Case-Based Reasoning (CBR) has the inborn ability of quantitative interpretation and prediction simply based on similarity of historical events even there is difficulty to get the working mechanism, principles, or build statistical models for those complex systems. Since it is a perfect model of human recognition and its working mechanism is very similar to the human primitive reasoning logic, the introduction of CBR to geographical system (Geo-case Based Reasoning) study to solve the problem of quantitative prediction was promoted. With tracking the front of international CBR research, expression and reasoning models of CBR have been promoted. And application of CBR model and method to specific geographical phenomenon has been fully discussed as well. The content of this paper can be summarized in the following aspects: (1) GeoCBR has been posed from the view of geographic case study. Furthermore, principle difference between GeoCBR and general CBR has been explored. (2) A comprehensive expression model for geographic case based reasoning has been established based on Tesseral model. (3) Geo-case based reasoning model was built in case of having fuzzy spatial-temporal distribution, a special FMP was used to calculate similarity of Geo-cases, and then reasoning model was given based on it. (4) With such a theoretical framework of Geo-CBR, a fishing ground prediction model, which can be regarded as a sample of Geo-CBR, has been built. This model has been applied to the East China Sea fishing center prediction. The validation result is satisfying.
出处
《地理学报》
EI
CSCD
北大核心
2002年第2期151-158,共8页
Acta Geographica Sinica
基金
国家863计划(818-11-03)
关键词
人工智能
地理案例推理
表达模式
推理模型
东海中心渔场
地学数据
artificial intelligent
Geo Case-Based Reasoning
representation model of Case-Based Reasoning
reasoning model
spatio-temporal distribution pattern
