节点文献
苹果病害事例库构建关键技术研究
The Research on Key Technology for Construction of Apple’s Disease Case Base
【作者】 邵世磊；
【导师】 周国民；
【作者基本信息】 中国农业科学院 ， 作物信息科学， 2013， 博士
【摘要】 苹果病害对苹果产量和品质都有较大的影响,我国在枰果病害的诊断与防治研究方面取得了大量的成果,也积累了丰富的诊断与防治经验,这些经验在生产管理中发挥了重要的作用。本文采用工智能中基于事例推理技术,以苹果病害诊断与防治工作的“经验”为研究对象,开展经验的事例化描述、检索与保存的关键技术研究。本文主要研究成果如下。1.针对苹果病害事例的特点,提出了一种适合苹果病害事例描述的事例库、事例索引库的结构,形成了完整的事例库体系结构,并系统地梳理了能够完整描述苹果病的属性集合。所设计的具有3级结构的索引表既面向事例也面向属性；既体现了每个属性在各个事例中的分布细节,也反映了每个属性在整个事例库中的存在价值,具有系统性、层次性和灵活性的特点。2.对索引表属性约简的关键技术进行了研究,提出了一种基于粗糙集理论和遗传算法的属性简约算法。该算法首先对取值为连续数值型的属性值进行了离散化处理,然后用粗糙集理论条件属性重要性度量方法来计算遗传算法中每代种群中各条染色体的适应度,最后用遗传算法实现属性约简。该算法充分考虑r属性在事例类别上的映射,能够借助遗传优化机制取得满意的约简属性子集。3.对苹果病害事例检索的关键技术进行了研究,提出了基于知识的事例检索方法。该方法首先利用索引词汇农和1级索引表对用户的查询进行了特征提取,其次根据特征的数量对事例库中各事例索引表中的异构属性进行同构化处理,然后采用基于致病因素知识的模板检索和归纳检索相结合的策略建立决策树,最后利用决策树实现事例初步匹配。这种方法提升了用户查询请求的典型特征对查询请求所发生背景的代表能力,缩小了事例检索的范围,提高了检索的效率和准确度。4.对苹果病害事例库维护和存储关键技术进行了研究,提出了一种基于事例推理性能的事例库更新机制和基于XML技术的存储结构。该机制将苹果病害诊断结果中的致病因素和病害类型作为判断事例推理性能的主要依据,通过比较新旧两个事例的推理性能,在增加和删除一个事例时对索引表进行调整。另外,针对农业生产特性和现代网络环境苹果病害事例库存储的要求,提出了一种基于XML技术的表达能力强、一致性好、安全性高的事例库存储结构。
【Abstract】 Apple’s disease has a greater impact on apple yield and quality. China’s agriculturaltechnology sector has made great progress in the application of disease diagnosisand prevention of theoretical and applied researches. A wealth of experience playsan important role in the application of disease diagnosis and prevention. However,subject to the cultural level of the majority of farmers and the development ofagricultural economy, dissemination, communication and use of these experiencesare greatly restricted. Along with the rapid development of information technology,adopting appropriate artificial intelligence techniques to describe, store and useexpert’s experience, it is of great significance to improve apple disease diagnosisaccuracy and control effect and improve the apple output and quality.Case-based reasoning technology is the only one in all artificial intelligencetechnology which is based on experience. This paper mainly researches the keytechnologies about ways to describe, retrieve and preserve the experience in theform of case. The study includes a total of four parts. Firstly, the paper designs thestructure of case base management system, the case base, the case and the indextable, and combes a set of attributes that describe apple’s diseases. Secondly, thepaper researches how to get a subset of attributes from the complete attribute setto construct the case base index table. Thirdly, the paper researches how to retrievea case which is the most similar to the background of the new apple disease problemfrom the case base, the solution of the retrieved case can offer the reference for thenew disease problem. Fourthly, the paper researches how to convert a diseaseproblem to a case, and the update mechanism of the case base when a new casejoins in it, meanwhile, the paper researches the storage structure of the case basebased on XML technology. Innovations in this paper are as follows:1.The research on the key technology on the design of case base structure, astructure suitable for apple’s disease cases description is given. In this paper, thestructure of case base management system, the case base, the case and the indextable are designed, and a complete set of attributes that describe examples of applediseases is combed. The index table is the core of the case base, and this paperdesigns3-grades index table structure based on the features of the apple diseasediagnosis and prevention work, in which the cases have different sets of attributes.The attribute-distribution-value in Grade1index table is used to describe thelocation of each attributes in the cases. The index structure is case-oriented andattribute-oriented，reflects the details of the distribution of each attribute in the casebase and the whole value of each attribute for the case base. The index structure issystematic, hierarchical and flexible.2. The research on the key technology on attributes reduction in the index table,a algorithm used for attribute reduction based on rough set theory and genetic results as the main basis to determine the case reasoning performance, and adjusts the index table when a case is added or deleted from the case base after comparing the old and new two cases’reasoning performance. In addition, a case base storage structure which is suitable for agricultural production characteristics and modern network environment, and has strong high security presentation, good consistency and high security is proposed.