知识不确定性度量的粒计算模型及其应用研究

Granular computing is currently a vivid research direction in the field of intelligent information processing. So far the work of granular computing theory, including uncertainty analysis of knowledge and mixed data processing, has not been sufficiently discussed yet. By using the tools of rough sets and fuzzy sets, the main objective of this project is to investigate uncertainty measure of knowledge and granular computing model, and apply them in mixed gene data classification. It will be realized through the following specific goals: (1) To explore theory and approach to uncertainty measure of knowledge, analysis the relations and differences between the proposed measures and the existing ones, and study the unified form of relevant uncertainty measures. (2) To develop the membership degree and the non-membership degree of intuitionistic fuzzy sets based on the upper and lower approximation set, and extend rough intuitionistic fuzzy set model. (3) To build the extended models of neighborhood rough set and random forest theories, investigate the evaluation criteria of gene subset, and present some corresponding hybrid gene selection algorithm. The results of this project will not only enrich granular computing theory by providing new theories and approaches for knowledge acquisition and uncertainty analysis in complex data, and will also be of value in application fields such as big data analytics and tumor diagnosis.

粒计算是智能信息处理领域非常活跃的研究方向。目前，该理论在知识不确定性分析、混合数据处理等方面尚未得到充分的研究。本课题以粗糙集、模糊集理论为工具，研究知识不确定性度量与粒计算模型，并将其应用于混合基因数据分类中。主要研究内容：提出知识不确定性度量方法，分析其与现有度量的联系和区别，研究相关度量方法的统一表达形式；基于上下近似集给出直觉模糊集的隶属度与非隶属度，建立粗糙直觉模糊集扩展模型；构建邻域粗糙集、随机森林等理论的扩展模型，研究基因子集评价准则，提出粒计算的混合型基因选择算法。本项目研究成果可为复杂系统的知识获取及不确定性分析提供新的理论和方法，并在大数据分析、肿瘤诊断等方面有广泛的应用价值。

不确定性问题是当前智能信息处理中的研究热点之一，粒计算是人工智能研究领域中模拟人类思维和解决复杂问题的新方法，是研究大规模复杂问题求解、大数据分析与挖掘、不确定性信息处理的有力工具。按照项目计划书的要求，项目组围绕知识不确定性度量的粒计算模型及其应用进行了系统而深入的研究工作，提出了条件信息粒度、直觉模糊集相似度、特征显著度等不确定性度量方法，建立了覆盖粗糙直觉模糊集模型和覆盖粗糙区间值直觉模糊集模型，构建了邻域粗糙集、模糊核聚类分析、随机森林等理论的扩展模型及其约简算法，设计了一些混合型基因选择算法，并将其应用于基因表达谱数据分析，在理论和应用方面取得了有价值的研究成果。这些研究工作不仅丰富了传统的粒计算方法，提升了粒计算处理大规模混合复杂数据的能力，而且对不确定性人工智能和基因表达谱数据挖掘的理论与应用的发展起到了积极的推动作用。