中文摘要：

目前关于流识别与分类的主流技术是基于统计学方法，其核心环节是提取有效的特征属性集，但这种方法的假设条件是，特征互不相关、数据也互不相关。正因为这种假设的不合理性，使得分类效果和识别性能有限，引入以数据相关性为核心的多重分形理论，从根本上摈弃独立假设的局限性与狭隘性，实现流的有效分类。为此，定义并论证流的分形谱，在此基础上推导流的估计谱，然后在定义的核域内基于灰色关联度进行估计谱分析，继而脱离特征提取过程实现流的分类识别。最后通过系列实验显示流的分形性和分形谱，并进行实际分类效果的纵向比较和横向比较。研究结果表明，基于多重分形理论的流分类识别方法，有效弥补了统计学方法所不可避免的独立假设缺陷，因此具有强大而高效的识别未知流的能力，也特别适合于动态多变的在线识别。

英文摘要：

The dominant methodology of flow identification and classification is based on statistical analysis, which mainly focuses on extracting efficient characteristics. However, its illogical hypothesis of characteristics independency and data independency limited the classification effectiveness. This paper introduced theory of multi-fractals to identify and classify flows in consideration of data dependency, which helped to fundamentally solve the problem of independent hypothesis. The work included the following aspects： defined the fractal spectrum of the flow, derived the fractal estimated spectrum, analyzed the estimated spectrum in core field by gray correlation, and then completed flows identification and classification without features extraction. Finally, series of experiments demonstrated the fractal nature and multi-fractal spectrum of flows, and made vertical comparison and horizontal comparison to express the effectiveness of actual classification. The result shows that flow classification with multi-fractals makes up for the defect of statistical methods caused by independent hypothesis, and also shows wonderful performance of this method when classifying unknown flows and classifying on line.