基于知识图谱的农业大数据碎片化知识发现方法研究

Knowledge discovery is a basic core problem and hot issue of the frontier in the field of artificial intelligence and agricultural informatization. How to solve the problem of multi-source heterogeneous, data overload, fragmentation of knowledge and so on has become an urgent demand of agricultural knowledge discovery in the era of big data. This topic puts forward a research on fragmentation of knowledge discovery method of big data in agriculture based on the mapping knowledge domain, its main research contents include: (1) Explore the suitable uniform expression and description system of fragmentation of knowledge and carry out the research on the method of characteristic representation and data collaboration preprocessing. (2) Use the knowledge model base, support vector machine (SVM), case based reasoning (CBR) and other theories or methods to study the fragmentation of knowledge mining and the knowledge storage management system based on mapping knowledge domain in the environment of big data. (3) In view of dynamic evolution of the big data, make a reaserch on the relevant fusion of fragmentation of knowledge, apply the deep forest algorithm to integrate the agricultural knowledge to realize the global knowledge of distributed autonomous agricultural knowledge updated in real time to realize the global distribution of distributed autonomous agricultural knowledge. (4) Construct the personalized user demand forecasting model based on mapping knowledge domain and establish a navigation model of agricultural knowledge to provide users with accurate, efficient and reliable agricultural knowledge services. It is of great theoretical and practical significance to further realize the automatic acquisition of agricultural knowledge and the intelligent seamless global development and integration of knowledge.

知识发现是人工智能与农业信息化领域的基本核心问题和前沿热点。如何解决分布自治式大数据环境下的多源异构、数据过载、知识碎片化等问题，已成为大数据时代农业知识发现的迫切需求。本课题提出基于知识图谱的碎片化农业大数据知识发现方法研究，主要研究内容包括：探索合适的碎片化知识统一表示与描述体系，开展农业大数据特征表示方法与数据协同预处理研究；拟采用知识模型库、支持向量机、案例推理等理论和方法，研究大数据环境下的碎片化知识挖掘以及基于知识图谱的知识优化存储管理；针对大数据的动态演化，研究碎片化知识的关联融合，拟采用深度森林算法进行农业知识融合，实现分布自治式局部农业知识的全局化融合；研究基于知识图谱的用户需求预测模型，建立个性化农业领域知识导航模型，为用户提供精准、高效、可靠的农业知识服务。本项目的研究成果，对进一步实现农业领域知识的自动化获取、智能无缝全局性知识演化与融合，具有重要的理论和实践意义。

农业知识发现是人工智能与农业信息化领域的基本核心问题和前沿热点。课题组利用知识图谱有效、直观的知识组织和表达形式、可推理、可扩展等特点，通过神经网络、主题图、主题模型、GOA和GBDT等人工智能方法，深入研究大数据环境下农业碎片化知识发现关键理论与方法，探索知识服务与应用，具体研究内容有：（1）大数据环境下农业数据协同集成方法；（2）基于知识图谱的碎片化农业大数据知识发现方法；（3）基于知识图谱的碎片化农业大数据知识融合；（4）面向个性化用户需求的农业大数据知识服务。以上研究内容为大数据环境下农业领域知识的自动化获取、智能无缝全局性知识演化与融合，提供有效方法，具有重要的理论与实践意义。