节点文献
基于GEP函数发现的智能模型库关键技术研究
The Research on Key Techniques in Intelligent Model Base Based on Function Finding by GEP
【作者】 元昌安;
【导师】 唐常杰;
【作者基本信息】 四川大学 , 计算机应用技术, 2006, 博士
【摘要】 计算机技术广泛而深入的应用对决策支持提出了更高的要求。智能决策支持系统(Intelligent Decision Support Systems,IDSS)应运而生。模型库子系统是决策支持系统的核心。但传统模型库的建立方法存在以下缺点:(1)需预先确定好模型库中模型的类型,DSS只能根据样本数据对现有的模型进行有关参数计算,进而让决策者根据结果来进行预测。这种模型库缺少了真正意义上的智能寻找模型类型的功能。(2)可能依赖领域的专家经验。如专家系统需要一个内容丰富而全面的知识库支撑,依靠的是某领域的专家经验,因此,该智能模型库对先验知识依赖性很强,难以实现真正意义上的智能化。(3)在预先不知模型类型时,需凭经验确定模型类型,含有主观和盲目因素。(4)传统方法目前不能支持复杂函数关系式和多分段函数关系式等模型的建立。(5)扩展性较差。针对不同的函数模型类型,在程序实现时就必须有一个新的程序模块。当模型库要进行扩充,系统就要为该模型增加新的代码。为了解决以上问题,本文将基因表达式编程(Gene Expression Programming, GEP)技术引入智能模型库的研究中。基因表达式编程具有极强的函数发现能力和很高的效率,并且在函数发现时不需要任何先验知识,无需预存函数模型的类型,避免了传统算法建模时事先选定函数类型的盲目性。
【Abstract】 With the rapid development of computer technology and application, Decision Support Systems (DSS) were demanded to deal with more complex problems.Intelligent Decision Support Systems (IDSS) have emerged as the times require. Sub-system of Model Base is core of DSS. The traditional methods of constructing Model Base have defects as follows.(1) Users must previously know the types of models in the Model Base. The DSS only calculates the unknown parameters of the models in Model Base in terms of the sample data. Then the decision-maker could predicts in terms of the models. These Model Bases depend on the selection of the type of the model, and do not implement intelligentization for Model Base.(2) The traditional intelligent expert Model Base depends on expert’s experiences in specific domain. Such Model Base needs an abundant and comprehensive knowledge base to support and depends on specialist experience of special fields. Consequently, this Model Base does not implement truly intelligentization.(3) If users have not transcendent knowledge for the data to be disposed, they will experientially determine the type of function model with subjectivity and blindness.(4) The traditional methods can not automatically discovery the complex functions and multi-section functions.(5) The traditional Model Bases are weak in expansibility. When the Model Base needs to be expanded, the system must be added new program code.
【Key words】 Gene Expression Programming; Intelligent Model Base; Decision Support Systems; Function Finding; Genetic Algorithm; Convergence; Attribute Reduction; Data Mining;