项目名称: 模糊认知集群优化的聚类算法
项目编号: No.61503306
项目类型: 青年科学基金项目
立项/批准年度: 2016
项目学科: 其他
项目作者: 袁锦锋
作者单位: 西交利物浦大学
项目金额: 21万元
中文摘要: 聚类分析是一项应用于多个领域的重要机器学习技术。常用的聚类方法在特征选择、运算效率、分类结果准确性及应用可行性方面均存在着一定的局限性。本研究计划针对以上问题提出了模糊认知集群优化的聚类算法(FCSOC)。FCSOC是一种使用网络模糊认知法和集群智能算法对聚类分析算法进行优化的混合方法。其中,网络模糊认知法是一种融合了模糊集和网络认知法的可以准确有效的量化专家意见及主观判断的决策方法。在FCSOC方法中,集群智能算法将被用于优化聚类算法,预计可提升聚类算法的准确性和计算效率。网络模糊认知法首先将被用于降低聚类方法的数据维度,预计可优化运算效率;其次将被用于量化数据集中的类别数据,预计可使聚类结果能真实反映用户需求。本计划提出的FCSOC方法在商业智能系统,社交网络分析以及模式识别领域的应用将被初步测试,并将在移动平台上进行应用的初步开发。
中文关键词: 聚类;认知模型;集群智能;模糊理论;最优化分析
英文摘要: Clustering analysis is an essential machine learning technique and applied in many application domains. There are various critical limitations in conventional clustering methods regarding feature selection, optimization efficiency,clustering result accuracy, and application feasibility. This proposal proposes the fuzzy cognitive swarm optimized clustering (FCSOC) methods to address these issues. FCSOC is the hybrid approaches of Fuzzy Cognitive Network Process (FCNP) and Swarm Intelligence (SI) for clustering analysis. The FCNP, the recent recognized decision method, combines fuzzy set and Primitive Cognitive Network Process to better reflect expert judgments. In FCSOC, Swarm intelligence is used to optimize clustering results to achieve expected accuracy and computational efficiency. FCNP is applied in FCSOC with not only feature selection reduced from high dimensions data, but also category data quantification, in order to achieve higher clustering accuracy and efficiency, as well as better reflections of users. The proposed hybrid FCSOC will be tested in business intelligence system, social network analysis, and pattern recognition. Some applications of FCSOC will be implemented in mobile platform.
英文关键词: Clustering;Cognitive Computing;Swarm Intelligence;Fuzzy theory;Optimization